ELECTRIC POWER RESEARCH INSTITUTE

Nondestructive Evaluation: Probabilistic Reliability Model for Thermally Aged Cast Austenitic Stainless Steel Piping

2012 TECHNICAL REPORT

Nondestructive Evaluation: Probabilistic Reliability Model for Thermally Aged Cast Austenitic Stainless Steel Piping

This document does NOT meet the requirements of 1OCFR50 Appendix B, 1OCFR Part 21, ANSI N45.2-1977 and/or the intent of ISO-9001 (1994).

EPRI Project Managers D. Kull T. Hardin

ItEEAIHINS-TI-TUTE

3420 Hillview Avenue Palo Alto, CA 94304-1338 USA PO Box 10412 Palo Alto, CA 94303-0813 USA 800 313.3774 650855 2121 -t
_____

1024966 Final Report, December 2012

DISCLAIMER OF WARRANTIES AND LIMITATION OF LIABILITIES

THIS DOCUMENT WAS PREPARED BY THE ORGANIZATION(S) NAMED BELOW AS AN ACCOUNT OF WORK SPONSORED OR COSPONSORED BY THE ELECTRIC POWER RESEARCH INSTITUTE, INC. (EPRI). NEITHER EPRI, ANY MEMBER OF EPRI, ANY COSPONSOR, THE ORGANIZATION(S) BELOW, NOR ANY PERSON ACTING ON BEHALF OF ANY OF THEM: (A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I) WITH RESPECT TO THE USE OF ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT, INCLUDING MERCHANTABILITY AND FITNESS FOR A PARTICULAR THAT SUCH USE DOES NOT INFRINGE ON OR INTERFERE WITH PRIVATELY OWNED PURPOSE, OR (11) THAT THIS DOCUMENT IS SUITABLE RIGHTS, INCLUDING ANY PARTY'S INTELLECTUAL PROPERTY, OR (111) TO ANY PARTICULAR USER'S CIRCUMSTANCE; OR (B)ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER (INCLUDING ANY CONSEQUENTIAL DAMAGES, EVEN IF EPRI OR ANY EPRI REPRESENTATIVE HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES) RESULTING FROM YOUR SELECTION OR USE OF THIS DOCUMENT OR ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT. REFERENCE HEREIN TO ANY SPECIFIC COMMERCIAL PRODUCT, PROCESS, OR SERVICE BY ITS TRADE NAME, TRADEMARK, MANUFACTURER, OR OTHERWISE, DOES NOT NECESSARILY CONSTITUTE OR IMPLY ITS ENDORSEMENT, RECOMMENDATION, OR FAVORING BY EPRI. THE FOLLOWING ORGANIZATION, UNDER CONTRACT TO EPRI, PREPARED THIS REPORT: Structural Integrity Associates, Inc.

THE TECHNICAL CONTENTS OF THIS DOCUMENT WERE NOT PREPARED IN ACCORDANCE WITH THE EPRI NUCLEAR QUALITY ASSURANCE PROGRAM MANUAL THAT FULFILLS THE REQUIREMENTS OF 10 CFR 50, APPENDIX B AND 10 CFR PART 21, ANSI N45.2-1977 AND/OR THE INTENT OF ISO-9001 (1994). USE OF THE CONTENTS OF THIS DOCUMENT IN NUCLEAR SAFETY OR NUCLEAR QUALITY APPLICATIONS REQUIRES ADDITIONAL ACTIONS BY USER PURSUANT TO THEIR INTERNAL PROCEDURES.

NOTE For further information about EPRI, call the EPRI Customer Assistance Center at 800.313.3774 or e-mail askepri@epri.com. Electric Power Research Institute, EPRI, and TOGETHER.. SHAPING THE FUTURE OF ELECTRICITY are registered service marks of the Electric Power Research Institute, Inc. Copyright 2012 Electric Power Research Institute, Inc. All rights reserved.

Acknowledgments

The following organization, under contract to the Electric Power Research Institute (EPRI), prepared this report: Structural Integrity Associates, Inc. 5215 Hellyer Avenue, Suite 210 San Jose, CA 95138 Principal Investigators D. Harris H. Qlan D. Dedhia N. Cofie T. Griesbach This report describes research sponsored by EPRI.

This publication is a corporate document that should be cited in the literature in the following manner: Nondestructive Evaluation: ProbabilisticReliability Model for Thermally Aged Cast Austenitic Stainless Steel Piping. EPRI, Palo Alto, CA: 2012. 1024966. 4 iii >

Product Description

A probabilistic fracture mechanics (PFM) method has been developed to support the evaluation of the effects of thermal aging embrittlement on the reliability of reactor coolant cast austenitic stainless steel (CASS) piping components. The probabilistic method relies on modeling of the contributing elements of a deterministic analysis (for example, geometry, loading conditions, materials, and known degradation effects), but it incorporates the uncertainties or unknowns in a probabilistic manner using distribution functions. The study described in this report presents the PFM model called cast austeniticstainless steelprobabilisticanalysis of reliability (CASSPAR) and the initial results of a probabilistic flaw tolerance approach that is proposed as a technical basis for managing the inspection and reliability of CASS piping in pressurized water reactors (PWRs). Background Thermal aging embrittlement of CASS piping is a time-limited aging effect that must be managed for plant license renewal. The methodology to be used is described in the Generic Aging Lessons Learned report under Aging Management Programs, Section XI.M12. An aging management program for CASS components must include a determination of the susceptibility of CASS components to thermal aging embrittlement based on casting method, molybdenum content, and percent ferrite. For potentially susceptible components, aging management is to be accomplished through enhanced volumetric examination or plant- or componentspecific flaw tolerance evaluation. Objectives An earlier study by the Electric Power Research Institute (EPRI) determined the allowable flaw depths in fully aged CASS piping. In that study, a deterministic fracture mechanics analysis was performed to determine the maximum allowable flaw depths. This analysis, which was performed using conservative treatment of input variables and the required American Society of Mechanical Engineers (ASME) safety factors, produced maximum allowable flaw depths that were small (approximately 10% of through-wall depth). Such flaws are too small to be reliably detected using current volumetric inspection techniques for CASS piping components. The objective of the study that is the subject of this report was to develop the allowable flaw depths in fully aged CASS piping using PFM methods.

Approach
An alternative to the conventional fracture mechanics method was developed. Rather than a deterministic approach with conservative inputs, the study's probabilistic approach accounts for the large amount of variability in materials and scatter in the correlations used to predict the thermal embrittlement. This approach bypasses the need for conservative bounding values and takes the scatter and uncertainty explicitly into account. The outcome of the analysis is the probability of failure (a component of risk) for a given crack size, rather than conservative estimates of allowable crack sizes with large safety margins. Results The CASSPAR model was developed to determine crack sizes (depth and circumferential extent) in CASS piping that would become unstable (fail) with a given probability when specified loads are applied. The probabilistic model predicts that crack sizes for the saturated material and unaged material are similar, with the aged material in some cases being more flaw-tolerant. At the level of 10-6 probability of failure, critical crack depths vary from 10% to 50% of the wall thickness, depending on the applied loads. The crack depths increase to 30-50% for loadings expected during normal operating conditions in cold leg primary piping in PWRs. Applications, Value, and Use The large flaw sizes predicted to have a 106 failure probability when normal service loads are applied suggest that CASS piping is quite flaw-tolerant. These results should be useful in developing ASME Code flaw acceptance standards for high-delta-ferrite materials. Keywords Aging management Cast austenitic stainless steel (CASS) Cast austenitic stainless steel piping Cast austenitic stainless steel probabilistic analysis of reliability (CASSPAR) Grade CF-8M Probabilistic fracture mechanics (PFM)

.<vi >

Abstract

This report presents a probabilistic fracture mechanics (PFM) methodology for determining the flaw tolerance of fully thermally aged cast austenitic stainless steel (CASS) piping components. A time-limited effect, thermal aging of CASS piping must be managed for plant license renewal through inspection or flaw tolerance analysis. When the effects of thermal aging embrittlement are potentially significant for piping based on high delta ferrite content (>20%), evaluation criteria are needed in order to use a flaw tolerance approach. Studies have shown that flaw tolerance methods can be used to evaluate reliability of piping systems by following the procedures in American Society of Mechanical Engineers (ASME) Section XI for flaw evaluation and acceptance. Appendix C of ASME Section Xl describes an evaluation of wrought stainless piping, welds, and CASS piping with ferrite content less than 20%, but high-delta-ferrite components are not covered. Much has been learned recently about CASS properties that can improve our knowledge of the materials' aging effects and flaw tolerance, making possible the development of flaw acceptance criteria and in-service inspection methods for CASS components similar to ASME Code evaluation methods for austenitic piping and welds. This report also presents a PFM method to evaluate the effects of thermal aging embrittlement on the reliability of Class 1 reactor coolant systems and primary pressure boundary CASS piping components. The PFM method relies on modeling, using key elements of a deterministic analysis and incorporating the inherent uncertainties in these parameters. The PFM approach has been used successfully for other aging effects in piping, such as intergranular stress corrosion cracking, and it is well suited for modeling and evaluating thermal aging in CASS piping. Presented herein are the PFM model and the results of a probabilistic flaw tolerance approach that provides a technical basis for management of thermal aging and reliability of CASS piping in pressurized water reactors. The results indicate that CASS piping components are highly flaw-tolerant, even in the thermally aged condition.

< vii >

Table of Contents

Section 1: Introduction ............................................ 1.1 Ba ckg ro und .......................................................

1.2 Methodology for Flaw Tolerance Analysis of C ASS Com ponents ........................................................... 1.3 Scope and Limitations of Study ............................. 1.4 Unit C onversions ................................................ 1.5 Symbols Used in This Report ................................ 1.6 Acronyms Used in This Report ..............................

1-1 1-4
1-5 1-5 1-6 1-6 1-9

Section 2: Basis for Fracture Mechanics Model .......... 2-1 2 .1 Determ inistic Basis .............................................. 2-1 2 .2 J-Integral Solution ................................................ 2-3 2.2.1Uncracked Relationships ...................................... 2-3 2.2.2 Fully Plastic Ji-Solution .......................................... 2-5 Section 3: Description of the Probabilistic Model ....... 3-1 3.1 Crack Sizes for a Given Failure Probability: Uncorrelated Toughness and Strength ................................. 3-1 3.2 Crack Sizes for a Given Failure Probability: Correlated Toughness and Strength .................................... 3-5 Section 4: Results for Example Problems .................. 4.2 Random Toughness, Tensile Properties, and A lloy Co ntent ................................................................... 4.3 Correlated Toughness and Flow Strength with Unce rta inty .................................................................... 4.5 Results with Delta Ferrite as Input ........................ 4.6 Tabular Presentation for Probability of Failure of 10,4 a nd 106.............................................................. 4-1 4 -4 4 -10 4-15 4 -16

Section 5: Discussion ............................................... 5-1 Section 6: Summary ................................................ Section 7: References .............................................. 6-1 7-1

Appendix A: Characterization of MaterialRelated Random Variables ..................... A-1 A. 1 Unaged Properties ........................................ A-1 A. 1.1 Unaged Tensile .......................................... A-1 A.1.2 UnagedToughness .................................. A-10
< ix >

A.2

Saturated Properties ......................................... Saturated Tensile, High Temperature ....... Saturated Toughness .................................... A.3 Comparisons with J-R Curve Data .................. A.4 Material Composition and Delta Ferrite Co nte nt ..................................................................... A.5 Degradation Parameter P ................................ A.6 Summary ........................................................ A.2.1 A.2.2

A-1 5 A-15 A-17 A-i 9 A-2 7 A-28 A-29

Appendix B: Toughness-Strength Correlations for Fully Aged CF-8M ............................... B-1 B. 1 Least Squares Fit ................................................. B-2 B.2 Predictor Intervals ............................................... B-3 B.3 Uncertainty ........................................................ B-4 B.4 Monte Carlo Simulation in CASSPAR .................... B-7 Appendix C: CASSPAR 1.0 Software User's M anual ................................................... C. I Objective ...................................................... C.2 Introduction .................................................. C.3 Required Input ................................................... C.3.1 Calculation Inputs .......................................... C.3.2 Input for Data Output ..................................... C.4 Calculation and Results ...................................... C.4.1 Calculation ................................................... C.4.2 Echo of Inputs ............................................... C.4.3 Results ..........................................................

C-1 C-1 C-1 C-2 C-2 C-3 C-3 C-3 C-3 C-4

List of Figures

Figure 1-1 Flaw disposition screening method for CASS p ip ing ma teria l ................................................................ 1-2 Figure 1-2 Critical and allowable flaw sizes from deterministic study (Elastic-plastic fracture mechanics [EPFM]: 8-ksi [55.2-MPa] membrane stress and 1O-ksi [68.9-M Pa] bending stress) ............................................... 1-3 Figure 1-3 Elements of flaw disposition in accordance with A SM E C ode Section Xl ..................................................... 1-4 Figure 2-1 A part-through, part-circumferential crack in a p ip e ............................................................................... 2 -1 Figure 2-2 A deterministic fracture mechanics model ................. 2-2 Figure 2-3 A fully plastic beam with combined tension and bend ing load ing .............................................................. 2-7 Figure 3-1 Procedures for evaluating crack sizes for a given failure probability: uncorrelated toughness and strength; D as a random variable .................................................... Figure 3-2 A probabilistic model identifying the locations of d e ta ils ............................................................................. Figure 3-3 Procedures for evaluating crack sizes for a given failure probability: correlated toughness and strength w ith epistem ic uncertainty .................................................

3-2 3 -3

3-6

Figure 4-1 Procedures for evaluating crack sizes that would fail with a given probability when defined loads are applied (Random cD, toughness, and tensile properties, a ll independ ent.) .............................................................. 4 -2 Figure 4-2 Critical crack sizes at various failure probabilities for the example problem with toughness as the only random variable .............................................................. 4-4 Figure 4-3 Critical crack sizes at various failure probabilities for example problem: random toughness and tensile properties, various loadings (Top: with 28-ksi [193-MPa] tension; center: with 8-ksi [55.2-MPa] tension and 20-ksi [1 37.9-MPa] bending; bottom: with 8-ksi [55.2-MPa] tension and 15-ksi [103.4-MPa] bending.) ..........................

4-7

<xi >

Figure 4-4 90% confidence intervals on probability for 0/n = 0.7, fully saturated, 8-ksi (55.2-MPa) tension, and 15-ksi (103.4-M Pa) bending .............................................

4-9

Figure 4-5 Results of Figure 4-3 with the 90% confidence interval shown for 0/i = 0.7, 10.6 probability (8-ksi [55.2-MPa] tension, 15-ksi [103.4-MPa] bending) ............. 4-10 Figure 4-6 Critical crack sizes at various failure probabilities considering correlation and uncertainty: 10th, 50th, and 90th epistemic percentiles (full saturated, with 8 ksi [55.2 MPa] of tension and 20 ksi [137.9 MPa] of be nd ing ) ....................................................................... Figure 4-7 Critical crack sizes at various failure probabilities considering correlation and uncertainty: 10th, 50th, and 90th epistemic percentiles (fully saturated, with 8 ksi [55.2 MPa] of tension and 15 ksi [103.4 MPa] of be nd ing ) .......................................................................

4 -12

4 -12

Figure 4-8 Critical crack sizes at various failure probabilities for example problem with 8-ksi (55.2-MPa) tension and 10-ksi (68.9-MPa) bending stress for fully aged CF-8M ....... 4-14 Figure 4-9 Probabilistic results of Figure 4-7 plotted with corresponding deterministic results from Figure 1-2 ............ 4-15 Figure 4-10 Critical crack sizes at 10-6 probability for various delta ferrite contents, 8-ksi (55.2-MPa) tension, and 15-ksi (103.4-M P) a bending ...................................

4-16

Figure A-i Data and fitted normal distribution of yield strength for unaged CF-8M at 650'F (343C) .................... A-2 Figure A-2 Data and fitted normal distribution of ultimate strength for unaged CF-8M at 650'F (343C) .................... A-3 Figure A-3 Data and fitted normal distribution of flow strength for unaged CF-8M at 650'F (343C) .................... A-3 Figure A-4 Data and relationship between a and flow stress for C F-8M ................................................................. A -5

Figure A-5 Stress-strain data for unaged and saturated CF8M, 290C (554F) (The open points and dashed lines are unaged; others are saturated.) .................................... A-7 Figure A-6 The a - ao,. data and fit for unaged and fully aged C F-8M ................................................................... Figure A-7 Comparison of data and fits for each of 10 specimens included in Table A-2 ....................................... < xii > A-8 A-9

Figure A-8 Data and quantiles of CF-8M unaged tensile tests included in Table A-2 .................................................... A-10 Figure A-9 Cumulative distribution of unaged CF-8M Charpy at room temperature (data from Table B of N UREG /CR-5385) .................................................... Figure A-I0 Figure II from NUREG/CR-4513 showing data points for high-temperature CR vs. room temperature Charpy impact energy ............................ Figure A-i1 Figure 12 of NUREG/CR-4513 showing relationship between exponent n and Charpy impact energy .....................................................................

A-i 2

A-i 3

A-13

Figure A-i12 Reproduction of Figure 21, page 50 of N UREG /CR-6 142 ...................................................... A-i5 Figure A-i 3 Normal probability plot of the flow stress ratio for the data of Figure A-1 2 with the normalizing parameter greater than 3 (Mean = 1. 189; standard deviation = 0 .07 1 .)...................................................... A -i 6 Figure A-i 4 Predicted quantiles of stress-strain for fully saturated CF-8M material at 290 0 C (553F) with data points from tensile tests identified in Table A-2 .............. A-i 7 Figure A-i 5 Figure 6 of NUREG/CR-4513 for CF-8M with nickel content less than 10% (C,,otis at room tem perature.) ........................................................... Figure A-i 6 Data of Figure 6 of NUREG/CR-4513 for CF-8M, Ni<1O%, along with median and fitted lOth and 90th percentiles .................................................. Figure A-i 7 Saturated elevated temperature J-R curves for CF-8M from NUREG/CR-4513 .......................................

A-1 8

A-I9 A-2I

Figure A-1 8 Test data and predicted quantiles of J-R curve for Heat 20 5 ................................................................ A -2 2 Figure A-1 9 Test data and predicted quantiles of J-R curve fo r Heat 75 .................................................................. A -2 3 Figure A-20 Test data and predicted quantiles of J-R curve fo r Heat 74 .................................................................. A -24 Figure A-21 Test data and predicted quantiles of J-R curve for Heat 7 5 8 ................................................................ A-25

Figure A-22 Test data and predicted quantiles of J-R curve fo r Hea t L..................................................................... A-2 6
< xiii >

Figure A-23 Distribution of CF-8M delta ferrite from data ........ A-28 Figure A-24 Distribution of (D from EPRI data, CF-8M ............. A-29 Figure B-1 The available data on high-temperature flow .......................... strength for a given room temperature Ca, Figure B-2 Available data along with selected percentiles as obtained from the least squares fit (mean) and the value of the standard deviation of errors ..................................... B-2

B-3

Figure B-3 The 10th, 50th, and 19th percentiles of distribution of aflo for a given Cvsat considering the small sample size obtained by the Mendenhall procedure ............. B-4 Figure B-4 The 10th, 50th, and 19th percentiles of distribution of ao for a given C,5o,considering the small sam ple size using W einberg ............................................. Figure B-5 Depiction of the end result ....................................... Figure C-1 Software interface ............................................

B-6 B-7 C-1

Figure C-2 Input tables: strength-toughness correlated (a) and uncorrelated (b) ....................................................... C -2 Figure C-3 The results table with correlation option "No". ......... C-3 Figure C-4 The Results table with correlation option "Yes". ........ C-3 Figure C-5 The Input Echo table ............................................. C-4 Figure C -6 Plot of results ........................................................ C -4 Figure C-7 An Epistemic Results worksheet .............................. C-5

4 xiv >

List of Tables

Table 1-1 Unit conversions ..................................................... 1-6

Table 2-1 Values of various parameters for tension and b e nd ing .......................................................................... 2 -4 Table 2-2 Values of/p, M , and Pas functions of ; for n = 2 , 5 , a nd 10 ............................................................. 2-5 Table 2-3 Sample portion of the table of h, (a, 0, 0.5, 10) n= 10 ; ;"= 0 .5 ................................................................. 2-8

Table 4-1 Values of a/h for various 0/it and selected probabilities: random tough ness-28-ksi (193-MPa) te n sion ............................................................................ 4 -3 Table 4-2 Values of a/h for various 0/7 and selected probabilities: random toughness, tensile properties and alloy content, and various stresses ...................................... Table 4-3 Summary of 90% confidence intervals for 0/n = 0.7, fully saturated, 8-ksi (55.2-MPa) tension, and 15-ksi (103.4-M Pa) bending .............................................

4-5

4-9

Table 4-4 Values of a/h for various and selected probabilities considering correlation and uncertainty (fully sa tura ted ) .............................................................. 4 -1 1 Table 4-5 Sample representative piping interface loads ........... 4-13 Table 4-6 Critical crack sizes for 32-in. (812.8-mm) line at the 106 probability level ................................................. Table 4-7 Critical crack sizes for 32-in. (81 2.8-mm) line at the 10-4 probability level ................................................. Table 4-8 Critical crack sizes for 12.75-in. (323.9-mm) line at the 106 probability level .............................................. Table 4-9 Critical crack sizes for 12.75-in. (323.9-mm) line at the 10.4 probability level .............................................. 4-17 4-17 4-18 4-18

Table A-1 Values of the mean and standard deviation of the tensile properties of unaged CF-8M at 650'F (343C) ........ A-2 Table A-2 Summary of 290'C (554F) stress-strain data available for CF-8M unaged and saturated ........................ A-6

<xv>

Table A-3 Summary of unaged charpy data for CF-8M at room tem perature ...................................................... Table A-4 NUREG data for CF-8M heats with fully aged, high-tem perature J-R curves .............................................

A-i 1 A-20

Table B-1 Summary of available data on room temperature CV,.t and high-temperature flow strength from same heats ...... B-1 Table B-2 Mean, standard deviation, and correlation coefficient for mean line fit ................................................ B-5

4 xvi >

Section 1: Introduction
The purpose of this report is to present a probabilistic fracture mechanics (PFM) methodology for determining the flaw tolerance of fiflly thermally embrittled cast austenitic stainless steel (CASS) piping components. The need for such a methodology stems from the fact that traditional deterministic fracture mechanics approaches could lead to very conservative -allowable flaw sizes that challenge the in-service inspection and management of these components for license renewal. Specifically, the first objective is to develop a probabilistically based flaw tolerance evaluation methodology for thermally aged CASS piping components to address license renewal issues. The second objective is to establish a methodology for determination of maximum allowable flaw sizes in CASS piping components at various probability-of-failure levels in order to assist in developing ultrasonic performance demonstration criteria for CASS piping components. Prolonged exposure of CASS to reactor coolant operating temperatures has been shown to lead to some degree of thermal aging embrittlement, which in turn leads to a loss of toughness and an increase in tensile strength with time. The magnitude of the reduction depends upon the type of casting method, the material chemistry, and the duration of exposure at operating temperatures conducive to the embrittlement process. Static castings are known to be more susceptible than centrifugal castings, high-molybdenum-content castings are more susceptible than low-molybdenum-content castings, high-delta-ferrite castings are more susceptible than low-delta-ferrite castings, and higher operating temperatures (320'C [610'F]) increase the embrittlement rate compared to the rate at lower operating temperatures (285C [550'F]). The extensive amount of fracture toughness data available for thermally aged CASS materials enables delta ferrite, molybdenum content, casting type, and service temperature history to be used as the bases for screening and evaluating components for continued operation during the license renewal term. Because of this thermal aging effect, utilities must evaluate the impact of loss of toughness using screening for susceptibility according to the General Aging Lessons Learned report [1]. For potentially susceptible CASS materials, a program must be implemented using either enhanced volumetric examination to detect and size cracks or plant- or component-specific flaw tolerance evaluation to show that the aging effects are managed. Currently, there are no acceptable approaches for analysis or inspection of high-delta-ferrite (>20%) CASS components, other than pump casings. One objective of the current study is to develop the basis for using flaw tolerance analysis for managing thermal

<1-1 >

aging of CASS piping. Figure 1-1 summarizes the flaw evaluation screening method for CASS materials, as drawn from Nondestructive Evaluation:Flaw Tolerance Evaluation of Thermally Aged CastAustenitic Stainless Steel Piping [2].

Figure 1-1 Flaw disposition screening method for CASS piping material [2]

< 1-2 )

A previous study of CASS piping integrity used deterministic fracture mechanics methods to evaluate flaw tolerance of the cold leg of a reactor coolant's primary system piping [2]. In that study, a flaw tolerance approach was used to evaluate both the critical and maximum allowable flaw sizes (that is, crack depth as a function of crack length) for typical cold leg piping of a pressurized water reactor (PWR) made from CF 8M cast material and having high delta content. It was determined from this study that using typical plant loading (stresses), the maximum allowable flaw depths-based on a conservative treatment of all input variables-were beyond the capabilities of current volumetric inspection techniques for CASS piping components, as illustrated in Figure 1-2. This creates a dilemma for utilities because it is not possible to demonstrate the absence of flaws that are smaller than the size that can be reliably detected by ultrasonic examination.
EPFM Solution for o,. = 8 ksi and oa = 10 ksi 1r r1

0.8

U
U.

0.6

-Fr---ii 'I

II II Ii

_____________________ _____________________ _____________________

_____________________

"E 0.4
I-

0.

IL

Ig~0 0.2

-57-J~-50.4

j5:7j

-I:*~:
0.6

_____

0.8

Length of Flaw, Fraction of Circumference (//2xr)

*20%

DFN (SFm = SFb = 1.0) * 20% DFN (SFm = 2.7, SFb = 2.3)

4 30% DFN (SFm = SFb = 1.0) 30% DFN (SFm = 2.7. SFb = 2.3)

DFN = delta ferrite number Figure 1-2 Criticaland allowable flaw sizes from deterministicstudy (Elastic-plasticfracture mechanics [EPFM]: 8-ksi [55.2-MPa] membrane stress and lO-ksi [68.9-MPa] bending stress) [2] An alternative to the conventional fracture mechanics method has been developed for the current study. Rather than a deterministic approach with conservative inputs, the study takes a probabilistic approach that accounts for the large amount of variability in materials and scatter in the correlations used to predict the thermal embrittlement. This approach bypasses the need for conservative bounding values and takes the scatter and uncertainty explicitly into account. The outcome of the analysis is the probability of failure (a component of

< 1-3

risk) for a given crack size, rather than conservative estimates of allowable crack sizes with large safety margins. When combined with the probability of a crack of a given size being present and the probability of detecting a crack as a function of its size, the overall failure probability can be evaluated and the benefits of inspection determined. PFM methodology lends itself to the analysis of systems where variability and uncertainties on the key parameters can be dealt with explicitly to calculate an overall probability of failure based on maintaining a safety goal. This approach offers the possibility of developing in-service inspection flaw acceptance criteria for CASS piping components, similar to American Society of Mechanical Engineers (ASME) Code Section X1, Subsections IVB 3500 and IWB-3600, for austenitic piping or dissimilar metal welds. This study develops a probabilistic fracture method for analyzing CASS piping materials and presents sample results performed for primary system piping in a PWR. 1.1 Background

ASME Section XI requires that all primary system pressure boundary components be inspected using a volumetric technique. When indications are detected during these inspections, evaluation procedures are provided in the Code for disposition of the indications, as shown in Figure 1-3. Therefore, the management of the integrity of system pressure boundary components is achieved through a combination of inspections and flaw evaluation. Although this same approach applies to CASS piping, certain limitations make direct application of this process a challenge.

InpetinsStandards (indications) I (Flaws) (Defects)

Replacement

Acceptable with No FurtherAction

Acceptable with Supplemental Reexamination

Figure 1-3 Elements of flaw disposition in accordance with ASME Code Section XI The first limitation has to do with the current state of the art of volumetric inspection of CASS piping. In practice, ultrasonic testing (UT) is the volumetric inspection technique normally used to inspect austenitic piping, components, and welds within the primary coolant piping system. The inhomogeneous coarsegrain structure of the statically or centrifugally cast CASS piping makes it difficult to inspect these materials using conventional UT techniques. Improved UT inspection techniques are still being developed for application to CASS piping materials [3, 4]. A key aspect of the inspections is the ability to detect flaws that are determined to be the maximum allowable depth or length to ensure that the piping system reliability can be maintained.

< 1-4>

The second limitation is that CASS materials experience thermal aging--or loss of material toughness-due to long-term exposure at the operating temperatures of PWRs. The management of this aging effect in CASS piping will require additional measures for plant license renewal, such as flaw tolerance analyses and inspections, to demonstrate that the integrity of the primary system is maintained. At present, there is no ASME Code-approved method for evaluating flaws in CASS piping with delta ferrite greater than 20%. The industry has been working on improved methods to evaluate and inspect CASS materials. The current study is a step toward developing an ASME Code method for evaluating CASS piping materials with high levels of delta ferrite. The current study uses published correlations that are not suitable for use with delta ferrite content above 25%, so the models developed herein are not applicable to these high-delta-ferrite materials. 1.2 Methodology for Flaw Tolerance Analysis of CASS Components Because of the associated inspection challenges, a proposed methodology to manage the structural integrity of CASS components relies on flaw tolerance evaluation. Flaw tolerance evaluation can be accomplished through a deterministic or probabilistic approach. Currently, the flaw acceptance criteria in ASME Code Section XI are all based on a deterministic approach using relatively lower-bound material properties. However, as described in NondestructiveEvaluation:Flaw ToleranceEvaluation of Thermally Aged Cast Austenitic Stainless Steel Piping [2], a deterministic approach with lower-bound material properties for flaw tolerance evaluation of CASS piping components could lead to very small allowable flaw sizes. This has also been demonstrated to a certain extent in previous studies [5-9]. The objective of this study, therefore, is the establishment of the PFM methodology for the flaw tolerance evaluation of the CASS piping components and not the acceptance criteria that will be addressed separately in the future. 1.3 Scope and Limitations of Study

As previously indicated, there are several grades of CASS materials, including CF3, CF3M, CF8, and CF-8M. However, the low-molybdenum CASS materials (CF3 and CF8) are the least susceptible to thermal aging effects. Furthermore, among the CASS materials, CF-SM is the most strongly affected by thermal embrittlement, so the present study will focus on this material. The methodology established in this report can be applied to other grades of CASS materials with the appropriate input parameters if the need arises. The following are the limitations on the study presented in the report: * As indicated, Grade CF-8M is the most susceptible to thermal embrittlement; therefore, the development of the methodology will concentrate on this grade of CASS piping.

< 1-5 >

Because the correlations employed are limited to ferrite levels lower than 25%, the methodology is limited to this ferrite level. The correlations employed are not applicable to alloy contents greater than 0.2 wt0/% niobium; therefore, the methodology is limited to this niobium content. " The correlations employed are limited to nickel content lower than 10 wt%. " The methodology is applicable to base metal. Associated welds will be addressed separately, such as through information in Effects of ThermalAging
on FractureToughness and Charpy-ImpactStrength ofStainless Steel Pipe Welds

[10].
1.4 Unit Conversions

Throughout this report, measurements are expressed as International System of Units (SI) units, English units, or both. Figure 1-4 sets forth equivalents that the reader can use to convert from one type of unit to the other. Table 1-1 Unit conversions
in. kip ksi or psi in.-kip
in.-lb/in2

mm N
MPaI

1 in. =

25.4 mm

I kip -448.2 N 1 ksi - 6.89 MPa;

s6.9Mo

1 psi = 0.00689 MPa

N-m
2 ki/m

I in.-dp - 112.99N-.m
1 in.-lb/in'
2 1 i/cm =

0.18 kJ/m

ftJb
ft-lb (for a normalized area of a Charpy-type specimen) OF
2 i/cm

If-b-1.36J
-

0.59 ft-lb

C CrF-32) x 5-9

1.5 a B
C1, C2, C3

Symbols Used in This Report crack depth a function constants in curve fit Equation 3-6 coefficient in crack growth resistance curve Charpy energy
median value of Charpy energy

CR Cv
Cvso

Cvai

fiuly aged (saturated) value of Charpy energy

41-6 >

do, d, D E F

constants in curve fit Equation 3-2 constant in Ramberg-Osgood plastic strain relationship modulus of elasticity tensile force pipe wall thickness dimensionless function in expression forJp moment of inertia value ofJ-integral plastic component of the J-integral stress intensity factor exponent in crack growth resistance curve bending moment on pipe cross-section dimensionless bending moment Ramberg-Osgood strain exponent probability density function estimated probability

h
hi

J
JP

K mn M

M
N P

tension force on pipe cross-section dimensionless tension force mean pipe radius outside pipe radius equals 1 if x is positive, -1 if x is negative a function x coordinate in linear curve fit distance from center of pipe y coordinate in linear curve fit a/h (or a constant in stress-strain relationship Equation 3-5) < 1-7 >

P
& Ro.
s(X)

T X y Y

00,

P3i

constants in linear curve fit

F 6c Aa C C(y)
E0
Emax

gamma function delta ferrite content crack extension plastic strain (or total strain in Equation 3-5) plastic strain across pipe cross-section strain at y=O maximum plastic strain in pipe cross-section mean and standard deviation of error standard deviation of estimate of standard deviation of error

.nin, -sd Esdsd

~X/(1+X)
0
K

half crack angular extent constant in strain relationship M/PRm

Eo/(ro+KRm)

k
Ia tU4 4

or shape parameter in lognormal distribution

fourth central moment

h/R

p G
-axial Gb

correlation coefficient normal stress

axial stress

bending stress
flow strength

OY1o

UyS, (P (D

yield strength angular coordinate around circumference of pipe material susceptibility

< 1-8 >

1.6 Acronyms Used in This Report

ASME CASS CASSPAR EPFM EPRI NUREG PFM PRAISE PWR UT American Society of Mechanical Engineers cast austenitic stainless steel cast austenitic stainless steel probabilistic analysis of reliability elastic-plastic fracture mechanics Electric Power Research Institute U.S. Nuclear Regulatory Commission report probabilistic fracture mechanics Piping Reliability Analysis Including Seismic Events pressurized water reactor ultrasonic testing

< 1-9 >

Section 2: Basis for Fracture Mechanics Model

The probabilistic analysis is based on a deterministic fracture mechanics model that is presented in this section, with some of the inputs treated as random variables and Monte Carlo simulation used to generate results. 2.1 Deterministic Basis

Due to Code inspection requirements and fracture considerations for piping systems, a circumferential crack in a pipe subjected to axial and bending loads is modeled, as shown in Figure 2-1.

Figure 2-1 A part-through,part-circumferentialcrack in a pipe The deterministic model to be used as a basis of the analysis is summarized in Figure 2-2.

< 2-1 )

C d l c IS i... z.
Figure 2-2 A deterministic fracture mechanics model

... .. ....

EPFM is employed because the toughness, strength levels, and applied loads are such that the use of linear elastic fracture mechanics is not valid (there is too much plasticity). The J-integral is used as the value of the crack-driving force. Under linear elastic conditions, J is equal to K 2/E, where K is the more familiar stress intensity factor. The value of the J-integral due to the applied loads is evaluated from the elastic and fully plastic J-solutions using the estimation procedure outlined in An EngineeringApproachfor Elastic-Plastic FractureAnalysis [11], which is also described in the second edition of FractureMechanics [12]. The elastic component ofJ is easily evaluated using corresponding stress intensity factor K-solutions. The K-solutions of the Ductile FractureHandbook [13] are employed here. The crack growth resistance (fracture toughness) is described by the JR curve, which is the value ofJ as a function of crack extension during a toughness test. The critical crack size is determined as the point at which the applied values ofJ and dJ/da exceed the material's resistance to crack growth. This approach is

< 2-2 )

commonly referred to as tearinginstability analysis [11, 12]. In a deterministic analysis such as the type currently employed in ASME Section XM flaw evaluation procedures, structural factors are applied to the material resistance curve (or loads) to determine the Code-allowable flaw size. 2.2 J-Integral Solution

No fully plastic J-solutions for both tension and bending for a part-through, part-circumferential crack are available for calculation of the applied value ofJ. The purpose here is not to generate new fully plastic J-solutions, which ideally would be evaluated by nonlinear, fully plastic finite element analysis. Instead, the purpose is to make maximum use of existing information, along with an interpolation scheme devised for estimating values of bi for combined loading. Information for tension or bending is available, such as from the Ductile Fracture Handbook [13], Cho et al. [14], and Y. Kim et al. [15], but usually for a very limited range of crack sizes. Due to lack of more complete information, the following approximate procedure was devised to estimate fully plastic J. 2.2.1 Uncracked Relationships The first step is to consider the plastic strain to be distributed linearly on the cross-section when no crack is present, which is expressed by Equation 2-1:
b(y) = -Fo ++ KY

Eq. 2-1

where y is the distance from the center of the pipe. This relationship is the standard assumption of plane cross-sections remaining plane. Fully plastic conditions are considered, which means that the elastic strains are negligible. The plastic strain-stress (a- - c ) relationship is written as

D)

=Eq.

2-2

This is the standard Ramberg-Osgood relationship with only the plastic strain; D and n are curve- fitting parameters that describe the relationship between stress and plastic strain. In the development of the fully plastic J-solution, the following dimensionless loading parameters are defined: M PR,, 1
__

,,,, -

-__ M

Eq.

2-3

mma

Co

+,CR,,I+

M+PRm

M is the applied bending moment, P is the applied tensile force, and R. is the mean pipe radius (the average of inner and outer radius). < 2-3 >

Using the strain relationship of Equation 2-1 with the stress-strain relationship of Equation 2-2 and integrating the stresses over the cross-section of the pipe provides the following usefiu relationships relating the strain to the tension force (P) and moment (M). These are expressed through the following relationship for the dimensionless loads P and M.

P(p, n)
m

max

J
0

I [p+(1-p)sino]'lu+(1-p)sinol"ndo Eq. 2-4

Ml(,u,n) =-

[p + I-u

i,0

u+(

R 2 Dhel,In

In

max

[0 (

i~d
Eq. 2-5

The function I is defined as 4X= - I x-+1 X< 0 x>0 E.2-6 Eq.2-

Combining Equations 2-3 through 2-5 shows that the following relationship also holds:

M(p, n)
n) + P(pt, )=2(pt, n) n) = M(2 M (p, n)

Eq. 2-7

The integrals in Equations 2-4 and 2-5 can be evaluated in closed form when Vis 0 and 1 (pure bending or pure tension). Table 2-1 summarizes the relationships between 4, X, ti, n, M, and P for tension and bending.

Table 2-1 Values of various parameters for tension and bending

pt

1 00 0

0 10

2x

0
2 [,-F I+A

< 2-4 >

The F term in Table 2-1 is the complete gamma function. For combined tension and bending (intermediate values of ), the integrals in Equations 2-4 and 2-5 are evaluated numerically. To ensure accuracy, Simpson's Rule with 105 intervals was employed. This defines the relationship between Pi and 4 for a given n. Table 2-2 provides a summary for n = 2, 5, and 10. Table 2-2 Values of p , ff and P as functions of ; for n = 2,5, and 10

ii
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.0000 0.8271 0.6988 0.6049 0.5389 0.5000 0.4811 0.4568 0.4287 0.3976 0.3641 0.3287 0.2920 0.2545 0.2166 0.1789 0.1415 0.1047 0.0688 0.0339 0.0000

P
6.2832 5.6984 5.1886 4.7401 4.3438 4.0000 3.6953 3.3805 3.0621 2.7444 2.4316 2.1270 1.8337 1.5536 1.2879 1.0372 0.8016 0.5809 0.3741 0.1809 0.0000

f
0.0000 0.2999 0.5765 0.8364 1 .0859 1.3333 1 .5836 1.8200 2.0413 2.2452 2.4313 2.5996 2.7504 2.8849 3.0047 3.1114 3.2062 3.2909 3.3667 3.4347 3.4961

/1 1.0000 0.6685 0.5354 0.4991 0.4928 0.4838 0.4710 0.4547 0.4352 0.4123 0.3860 0.3564 0.3237 0.2881 0.2501 0.2102 0.1689 0.1268 0.0844 0.0420 0.0000 6.2832 5.7348 5.3069 4.9683 4.6652 4.3562 4.0383 3.7130 3.3824 3.0486 2.7145 2.3833 2.0589 1.7447 1.4438 1.1588 0.8912 0.6418 0.4105 0.1968 0.0000

M 0.0000 0.3018 0.5896 0.8766 1.1659 1 .4517 1.7306 1 .9989 2.2546 2.4940 2.7141 2.9126 3.0880 3.2398 3.3688 3.4763 3.5645 3.6360 3.6932 3.7386 3.7744 1.0000 0.5459 0.4986 0.4943 0.4900 0.4811 0.4699 0.4555 0.4384 0.4179 0.3940 0.3666 0.3355 0.3011 0.2634 0.2231 0.1805 0.1364 0.0912 0.0456 0.0000

P
6.2832 5.7845 5.4460 5.1369 4.8278 4.5082 4.1829 3.8506 3.5132 3.1718 2.8289 2.4874 2.1511 1 .8235 1.5087 1.2096 0.9284 0.6669 0.4252 0.2033 0.0000

M i
0.0000 0.3044 0.6048 0.9059 1 .2069 1.5026 1 .7927 2.0729 2.3419 2.5948 2.8284 3.0398 3.2261 3.3863 3.5198 3.6282 3.7135 3.7783 3.8263 3.8600 3.8825

2.2.2 Fully Plastic J-Solution

Relationships for the fully plasticJ for pure tension and pure bending are provided by Cho et al. [14]. The intermediate functions Tand B are employed.

T =2cos-'(AsinO)7r

2-

0 (2-2++a)
4

Eq. 2-8

2[l + (2asa/h h R, 1 Rm+ 1 h 2

-[1-)(2

2- + a) +1- + a 2]

Eq. 2-9 Eq. 2-10

Eq. 2-11

< 2-5 >

sinO

q 2-12

The following expression for the fully plastic Jwith combined tension and bending is employed here. M is the applied bending moment, and P is the applied tension force (which includes the pressure-induced axial stress).

JP
-a)h4

(M
2

)2

TI
Eq. 2-13

B)

This can be written in dimensionless form as follows:

n+1

JP
_______

(I

a)h 1 (a, 0

,n

{~a[~4

+-T

Eq.

2-14

Equation 2-13 reduces to the corresponding expressions in Cho et al. [14] for pure tension and pure bending. The function hi plays a significant role in Equations 2-13 and 2-14, and the fully plastic J-solution consists of a table of this function as a function of crack size (a

and 0), mixture of tension and bending (c), and strain hardening exponent (n).
As explained previously, tabulated values of hi are very sparse. Of the available solutions [13-15], the widest range of crack sizes is given by the Electric Power Research Institute (EPRI) [13], but only for Rm/h of 10, and the bending results in the Ductile FractureHandbook [13] are very limited. Because of the strong need for a wide range of crack sizes (and less need for a range of Rm/h), the Zahoor tension solution is employed here, along with the following procedure for estimating the mixed tension and bending hi. It is assumed that the dimensionless JP of Equation 2-14 for combined tension and bending is the same as for pure tension. In other words, it is assumed that for a given maximum plastic strain (and tensile properties and crack size), the value of the plastic component ofJis independent of whether this strain is imposed by tension loading, bending loading, or a mixture of tension and bending. This is depicted for a beam in Figure 2-3.

< 2-6 >

mixture Figure 2-3 A fully plastic beam with combined tension and bending loading This would appear to overestimate the value ofJ for combined loading, but it is less conservative than assuming the loading to be totally tension. The assumption that the dimensionlessJ is the same for a given maximum plastic strain allows the tension and bending values of hi to be obtained from the tension-only solution. Noting that for pure tension (=0, P = 27r, and MQ = 0), the following relationship between the hi for tension and combined loading is obtained with Equation 2-14:

( )T
h 1 (a,0,0, n)= h, a,0

Z ) M(
~,0
2

-1

+
1
3('; 1) 4B(x, 0)

+rP( ,)
Eq. 2-15

4B(a,O0)

TiT(a, 0)

111

The values of hi(aO,O,n) are for tension and are obtained from EPRI [13]. Equation 2-15 then allows the values of bi(a,O,C,n) to be evaluated for combined loading (" other than 0). Table 2-2 summarizes the functions M(,;,n) and

Equation 2-15 was used to provide a table of hi(a,O,,n) for all crack sizes included in the tension table of the Ductile FractureHandbook. Therefore, the table is limited to Rm/h =10, n=2, 5, and 10. Values of a from 0.1 to 0.8 in increments of about 0.1, 0/1r from 0.05 to 1.0 in increments of 0.05, and from 0 (tension) to 1 (bending) in increments of 0.05 are included. It is desired to provide results beyond a of 0.8, so the fully plasticJwas extrapolated linearly out to a of 0.95 and the corresponding values of bhincluded in the table. Interpolation in the hi table is used in the computation of the fully plastic J. The table itself is large and incorporated in the software used to generate numerical results. Table 2-3 is a sample for a given combination of loads (L= 1/2) and strain hardening exponent (n = 10). This is just a sample of the much larger table used in the software described in AppendLx C. As previously covered, the failure criteria are that crack instability occurs at a crack size (or load) where the applied value ofJ and dJ/da are both higher than the values from the crack growth resistance curve (see Section 3.1). This is referred to as tearinginstability [11-13]. The applied value off includes both the plastically corrected elastic component and the plastic component from Equation 2-13. 4 2-7 >

Table 2-3 Sample portiot7 of the table of ho, (a, 8, 0.5, 10) (n'= 10;

4 =0.5)
10.40
9.4907 10.3090 10.6805 12.5 254 14.4307 15.497 3 15.9406 17.2579 19.976 1 24.3032 21.6852 23.2 257 19.9662 16.5474 13.7677 12.9046 11.785 1

I a 10.05
0.100

10.10

10.15

10.20

J0.25 10.30
10.6306 12.5143

10.35
9.9871

J0.45 10.50
9.0349 9.8 126 10.063 11.446 12.980 13.660 13.695 14.343 16. 388 19.804 16.808 17.608 14.545 11.830 9.751 8.3 58 7.189 8.8446 9. 3912 9.724 11 10.705 5 12.0583 12.3409

10.55
8.7063 9.1187 9.45 26

0.60 10.659
8.6685 8.9814 9.3492 10.019 11.028 8.7261 8.9682 9.4018 10.032 11.001 10.936 10.8 16 10.659 11.3 73 12.506 10.73 1 10.736 8. 9095 7.0159 5. 5615 3.0652 1.9746 1. 1655 0.6302 0.3188 0.1368 0.0530 0.0176 0.0094

0.70 1
8. 8471 9.0877 9.6013 10.252 11.214 11.212 11.227 11.152 11.809 12.777 11.242 11.2 71 9.680 8.104

0.80 0.75
91.0501 9.3 139 9.9336 9.3476 9.6874 10.383

0.86
9.8171 10.154 10.917

0.90

10.95

11.00

1 10.5355 12.4827 12.5680 11.9850 11.2801

10.0328 10.5032 10.9387 10.7036 11.5788 13.0479 14.4716 15.2139 16.6277 17.5311 19.0665 20.8556 21.3421 22.9255 23.3736 23.6406 22.0261 35.0790 20.4882 10.0227 4.0911 1.3487 0.3363 0.0562 0.0050 0.0010 11.3208 12.3695 14.1596 15.7963 16.9298 18.7196 20.3476 22.3039 24.8068 26.4665 28.4561 30.6483 33.3566 11.9984 03.2464 05.3766 17.2123 18.8569 21.0643 23.6137 26.0581 29.4830 32.6645 35.1967 39.7600 45.7439

0.125
0.150 0.200 0.250 0.300 0.350 0.400 0.450 0.500 0.550 0.600 0.650 0.700 0.750 0.800

112.9526 14.9989 15.4462 14.3424 13.1514 12.1115 110.169

114.377517.4629 17.1532 16.4248 13.7407

117.5164 22.3368 22.5822 20.5384 18.1110

1 1.480

15.8523 14.174 16.3 30 18.038

10.2360 11.3034

10.67610 11.288 12.085 11.704 1 12.356 13.334 13.824

119.8297 25.6449 25.7927 24.2304 21.5056 18.7416 124.5123 33.6392 32.1606 29.2703 25.7315 126.0645 38.6740 1 31.6544 45.0176 1 37.3713 21.5529

11.4934 11.020

11.7810 12.659 11.99 17 1 1.9904 02.6700

39.6269 37.0362 29.1305 23.2248 19.053 45.2638 44.9838 33.0751 26.1505 2 1.1071

12.10467 11.2295 10.828 12.4429 13.9676 16.1929 13.6951 14.0596 11.5816 9.4007 7.5615 5.9561 4 .80084 3 .7068 2. 7629 1.9996 1.4140 0.9861 0.6907 0.5860 11.3104 12.769 7 10.789 11.535

13.252 14.887 13.348 04.141 15.235 16.265 17.792 17.450 18.539 17.936 17.210 14.479 19.94 11.74 5.829 2.434 0.830 8.217 0.039 0.0040 0.0009

50.6746 49.0212

53.4037 38.3080 31.6721 24.613 38.2484 39.4609 39.6732 37.3518 37.2294 35.1342 30.066 28.9 59 31.2 29 28.6811 24 .265 2 1.064

1 44.0884 59.8684 66.2498 59.5081 48.6112 149.5111 66.1641 70.1556 64.5300 52.4091

13.9862 12.827 11.9206 11.049 11.9610 9.9309 7.7066 6.5797 4.5851 3.42 76 2.4224 1.6286 1.0452 0.64 19 0.3787 0.2 170 0.1639 10.938 9.061 6.814 6.132 3.749 2.593 1.67 1 1.006 0.566 0.2966 0. 1437 0.064 1 0. 04 17

13.7803 15.383 12.4 563 12.8843 11.3 112 9.4152 14.479 15.147 14.076 11.388 9.171 10.682 6.391 3.262 1.409 0.513 0.1469 0.0304 0.0038 0.0000

I55-5838 75.9391 72.8515 68.3761 54.1266

169.7911 85.2074 85.4421 72.5296 55.9939 1964.2798 100.513 1 103.604 118.615 1131.590 87.7026 79.1800 56.3176 55.5888

107.1061 98.7024

4.9948 6.2249
3.0022 1.8562 1.0269 0.5074 0.2210 0.0827 0.0255 0.0061 0. 0026 4.9892 3.02 56 1. 5990 0.7355 0.2889 0.0933 0.02 32 0.0039 0.00 13

36.149S564.5171 56.2520 32.6421 15.8279 6.3816 2.0674 0.5025 0.0888 0.0067 0.0013 82.7550 47.8195 23.0614 9.23044 2.9609 0.7102 0.1122 0.0091 0.0017

146.244 136.1148 95.5219 58.0791 35.3243 2 1.008 160.626 149.6678 98.9859 58.9075 34.8124 59.9436 34.226 33.736 33.537 20.129

0.820 1146.642
0.840 0.860 0.880 1 165.185

178.114 165.3965 103.342

19.0646 10.5447 5.998 17.9612 16.9483 9.3 112 8.1721 4. 8927 3.92 78 3.1282 2.50 11 2.0520 1.9020

1 188.743 200.125 184.5025 109.103 61.4558 1219.855 229.006 208.943 242.199 117.085 128.726 63.8314

0.900 1 263.086 268.975

0.920 0.940 0.950

67.7110 33.9D6

16.1593 7.1914 15.7776 6.4264

1 327.566 328.448 291.177 146.864 74.31931 35.3263 1 434.590 427.054 371.937 178.083 86.45451 38.8529 1 520.026 505.745 436.265 203.526 96.6784 42.1828

16.15399 5. 9758 16.86906 5. 92 97

< 2-8 >

Section 3: Description of the Probabilistic Model

The deterministic approach described in Section 2 provides the basis of the probabilistic model. Rather than a deterministic approach with conservative inputs, a probabilistic approach is used that accounts for the large amount of variability in materials and scatter in the correlations used to predict the thermal embrittlement. Such an approach bypasses the need for conservative bounding values and takes the scatter explicitly into account. The outcome of the analysis is the probability of failure (a component of risk) for a given crack size, rather than conservative estimates of allowable crack sizes with large safety margins. Similar analyses of piping reliability, such as the Piping Reliability Analysis Including Seismic Events (PRAISE) code [16], have been performed. PRAISE considers similar problems of piping reliability resulting from the growth and instability of initial cracks in pipes, but it considers tensile properties and material toughness to be deterministically defined. The material toughness is defined by the crack growth resistance curve, which is the crack extension as a function of applied J during a test. The crack growth resistance curve for cast austenitic materials is subject to significant scatter, even for a given composition, temperature, and delta ferrite content [17]. The toughness decreases with time, transitioning from an unaged condition to a fully embrittled condition (saturated) over a period, with the rate of transition depending on temperature and composition. Under typical reactor operating conditions, the fully saturated condition is reached in about five years (-40,000 hours). This transition is not treated in this report, and this analysis considers only unaged and filly saturated conditions. Only CF-8M material was considered in this model because its behavior is the poorest among the cast austenitic materials commonly used in commercial nuclear power plants (CF3, CF8, CF-8A, and CF-8M) [17]. It has been observed that the tensile strength increases and the ductility decreases with age [18]. 3.1 Crack Sizes for a Given Failure Probability: Uncorrelated Toughness and Strength Figure 3-1 summarizes the probabilistic model for obtaining failure probability for a given crack size (or, -alternatively, the sizes of cracks that have a given failure probability) when the strength and toughness are considered to be independent. Section 3.2 discusses the model if correlations are considered because more < 3-1 >

evidence was gathered for the strength-vs.-toughness correlation. Figure 3-1 is intended to provide an overview of the procedure involved, and details of the information in the boxes are provided in Figure 3-2.

L
I,
Iv

Jr

~4i~

4,

Figure 3-1 Procedures for evaluating crack sizes for a given failure probability: uncorrelated toughness and strength; (P as a random variable

( 3-2 >

Figure A-24

Figure A-3 Unagedc

Figure A-1 3

Dis~triution of C,,.

Dis Of Unaged C, m

Cx - 0,

+
Figure A-16 6

Figure A-6

Eqs. A-8a, 9a,

2.2.2

IT.
Value oftJ .

Only needed

if analyzing unaged material

t.4i

Figure 3-2 A probabilisticmodel identifying the locations of detoils As depicted in Figure 3-1, the following material-related random variables are considered: " " " The initial Charpy impact energy (unaged Cv) The initial flow strength (unaged on,) The ratio of fully aged flow strength to unaged flow strength
(Ono sat/Ono unaged)

<3-3>

" "

The fully saturated Cv (Cvs, ) (for a given D) The material's susceptibility, as expressed by the parameter D, calculated as follows: Eq. 3-1

( = dc (Ni + Si +Mn)2 (C + O.4N) / 5

Numerical results are generated by the probabilistic model by straightforward Monte Carlo simulation. The circumferential crack length is fixed, and values of the input random variables are sampled from their respective distributions. For each set of samples, a critical crack depth is calculated. This is done repetitively to generate a set of critical crack depths, which defines the distribution of these depths. The crack depths at the desired quantiles are then obtained from this distribution. The procedure is then repeated for the next circumferential crack length. In this section, all random variables are considered to be well defined, that is, there is no uncertainty due to lack of data. As reviewed in Appendix A, a large quantity of data is available. Accordingly, only aleatoric uncertainty is considered in this section. Epistemic uncertainty is considered in Section 3.2 and Appendix C. Data for static and centrifugally cast materials are combined in order to increase the size of available data sets. Figure 3-1 provides an overview of the probabilistic model. Figure 3-2 provides additional details, including identification of sections of this report where information on the distributions can be found. The distributions of the random variables and the data upon which the distributions are based are described in Appendices A and B. Random saturated flow strength and crack growth resistance curve properties are derived from the random variables (Dand unaged anmthrough relationships described in Appendix A, as drawn from two U.S. Nuclear Regulatory Commission reports (NUREGs) [17, 18]. These relationships include the following:
3 loglo C,,s., = do + d,e-3

Eq. 3-2

J = {404(25.4),- (C,

)0.41

[Aa]m

Eq. 3-3 Eq. 3-4

mn = 0.244 + 0.06log, (C,.)

6 = a + Cc
E
a
C1 .-o - C

(
o

'I

Eq. 3-5

a'flo

Eq. 3-6
3

< 3-4)0-

Cvsa, is lognormally distributed, with Equation 3-2 providing the median. The shape parameter (y of Equation A-6 in Appendix A) is taken to be the same as evaluated for unaged material. (Figure A-16 shows that this provides a reasonable representation of the aged Charpy data.) See Equation A-4 for values of cl, c2, and c3. The loads in the model can be random or deterministically defined. Similarly, the value of the degradation parameter (D can be defined, or the distribution defined from a number of plants can be used (see Section A.5). If the loads and inputs for a specific location in a specific plant are known, results can be generated for that location. Alternatively, load and 0 distributions representative of a fleet of plants can be used to generate failure probabilities for the fleet. 3.2 Crack Sizes for a Given Failure Probability: Correlated Toughness and Strength Section 3.1 considered the toughness and strength to be independent, but it is likely that these two random variables are correlated. Heats that suffer a greater loss of toughness might be likely to have a larger increase in strength (and loss of ductility). (According to the Atlas Foundry glossary, a heat is "a single charge of metal to be used for pouring into mold cavities.") Data on fully saturated toughness and strength from the same heat of material are needed to quantify this, and such data are very limited. Uncertainty in predictions resulting from a lack of data should be included in the results-this is where epistemic uncertainty enters. Figure 3-3 depicts the procedure employed when correlations between saturated toughness and strength are considered. In this case, (D is the governing input variable, which controls the distribution of the Cv... The fully saturated flow strength (Clno) is then derived from the Cv and the correlation between the impact energy and flow strength. Appendix A (see Section A.5) summarizes the data for CVsa, and aflo and describes the correlation and treatment of uncertainty from the lack of data. A treatment of uncertainty resulting from the limited data is provided, which is preliminary and could change as more rigorous approaches are applied. The scatter in the relationship between Cw,, and at,, can be the result of inherent randomness (aleatoric) and lack of data (epistemic). The article "Aleatoric or Epistemic? Does it Matter?" [19] provides a discussion, and the epistemic and aleatoric contributions to the probabilities to be presented are separated.

<3-5 >

LL

Figure 3-3 Procedures for evaluating crack sizes for a given failure probability: correlated toughness and strength with epistemic uncertainty

4 3-6 >

Section 4: Results for Example Problems

Results are presented for example problems of a part-circumferential crack in a pipe. Unless otherwise stated, the problems are for a cold-leg-to-pressure-vessel joint. The pipe dimensions are an outer radius (Ro) of 16 in. (406.4 mm) and a thickness (b) of 2.25 in. (57.2 mm). The pressure is 2250 psi (15.5 MPa), which imposes axial pressure stress of approximately 8 ksi (55.2 MPa). Different magnitudes with various combinations of tension and bending will be considered. In all cases, the stresses employed use loads based on elastic calculations. AppliedJuses a tension force, P, and bending moment, M, as seen in Equation 2-13. The tension stress reported includes the pressure-induced axial stress and is related to the tension force through Equation 4-1.
P
= a tension
2

;r Rmh

Eq. 4-1

Similarly, the reported bending stress is related to applied bending moment by the relationship
M
= u

bendingI

Eq. 4-2

Ro where I is the moment of inertia. The nominal temperature considered is 550'F (288-C). Figure 4-1 depicts the procedures for generating results for random tensile properties, the degradation parameter (0), and toughness. Other conditions are covered in the following. Conventional Monte Carlo simulation is employed using the random number generator in Visual Basic for Applications programming.

<4-1 >

-----------------........................ .................

Deterministic

FM Calculation of Critical
Crack Size

......................

i ......

........................................

.........

Processing of Monte Carlo Results

Figure 4-1 Procedures for evaluating crock sizes that would fail with a given probability when defined loads are applied (Random (D, toughness, and tensile properties, all independent.) 4.1 Random Toughness

As described in Appendix A, there are two types of random distributions in terms of toughness. One is the random distribution of toughness for a fixed chemical composition, and the other is the random distribution of toughness with respect to the distribution of chemical compositions. In this section, the only random variable is the toughness for a fixed chemical composition. The composition, delta ferrite content, and tensile properties are based on an actual material-Heat 205 of NUREG/CR-4513 [17]-and are summarized as follows:
" Cr = 17.88 wt%

* "

Mo =3.34wt% Si=0.63wt% <4-2 >

0 " "
"

Ni= 8.80 wt%/6 Mn=0.93wt% C=0.04wt%/o 6, = 15.9 oy= 29.0 ksi (199.9 MPa)
*

"
" " "

57.2 ksi (394.4 MPa)

a = 50.14 n = 6.408

Note: a is as in Equation 3-6; n is as in Equation 3-5. Analyses were run for unaged and fully aged properties, with the results summarized in Table 4-1 and plotted in Figure 4-2. The elastically calculated stress is 28 ksi (193 MPa), considered to be all tension. The number of trials used in the Monte Carlo simulation was 5x10 6 . Table 4-1 Values of a/h for various O/7 and selected probabilities:random toughness- 28-ksi (1 93-MPa) tension

10W 10,

0.62 0.65 0.71 0.73 0.76 0.70 0.73 0.76 0.78 0.80

0.53 0.56 0.61 0.66 0.68 0.61 0.65 0.68 0.70 0.73

0.49 0.52 0.56 0.59 0.63 0.55 0.59 0.63 0.65 0.68

0.46 0.48 0.50 0.53 0.57 0.50 0.53 0.56 0.58 0.60

0.44 0.45 0.47 0.49 0.50 0.47 0.48 0.50 0.53 0.56

0.43 0.44 0.46 0.47 0.49 0.46 0.47 0.49 0.50 0.53

SaturadW

0.01 0.1 0.5 106 104

Unaged

0.01 0.1 0.5

<4-3 >

1 -- U0.8
-41-

5.E-01
1.E-01

1.E-02 5
0.4U
- -

1.E-04
1.E-06

-5.E-01 -1.E-01

0.4

.E-02

0.2

1.E-04
1.E-06

0II

0.2

0.4

0.6

0.8

Ohr
Note: Solid lines are unaged; dashed lines are aged (28-ksi [193-MPa] tension)

Figure 4-2 Critical crack sizes at various failure probabilitiesfor the example problem with toughness as the only random variable

4.2 Random Toughness, Tensile Properties, and Alloy Content

The effect of random material properties was further investigated by considering the randomness of the tensile properties and alloy content. The distribution of thermally aged and unaged tensile properties and toughness are described in detail in Appendix A. Analyses were performed for unaged and fully aged properties, with the results summarized in Table 4-2 and plotted in Figure 4-3. Various stress levels consisting of a combination of tensile and bending components were considered. The number of trials in the Monte Carlo simulation was 5x10 6 for all cases, unless otherwise stated.

<4-4>

Table 4-2 Values of a/h for various 017r and selected probabilities:random toughness, tensile properties and alloy content, and various stresses

10" 107 Saturated 0.01

0.10 0.24 0.48

0.10 0.22 0.43

0.10 0.21 0.41

0.10 0.21 0.39

0.10 0.20 0.37

0.10 0.20 0.36

0.1
0.5 107
101

0.73
0.95 0.10 0.15

0.65
0.81 0.10 0.14

0.58
0.76 0.10 0.13

0.52
0.68 0.10 0.13

0.48
0.62 0.10 0.13

0.47
0.59 0.10 0.14

Umagd

0.01
0.1

0.42
0.65

0.38
0.55

0.36
0.51

0.35
0.47

0.33
0.44

0.33
0.43

0.5
1(0'

0.84
0.30 0.62 0.95 0.95 0.95 0.32 0.49 0.95 0.95 0.95

0.76
0.29 0.54 0.83 0.95 0.95 0.30 0.43 0.80 0.95 0.95

0.70
0.28 0.50 0.77 0.95 0.95 0.29 0.42 0.75 0.94 0.95

0.63
0.26 0.46 0.69 0.79 0.91 0.27 0.39 0.67 0.78 0.88

0.58
0.26 0.44 0.63 0.72 0.79 0.27 0.37 0.60 0.70 0.78

0.55
0.26 0.43 0.59 0.68 0.76 0.27 0.37 0.57 0.66 0.73

28-ksi tension

107
Saturated 0.01 0.1 0.5 10,

10F
Unaged 0.01 0.1 0.5

< 4-5 >

Table 4-2 (Continued) Values of a/h for various O/ir and selected probabilities:random toughness, tensile properties and alloy content, and various stresses

10 10 Saturated 0.01 0.1 0.5 10r 104 Unaged 0.01 0.1 0.5

0.50 0.88 0.95 0.95 0.95 0.56 0.78 0.95 0.95 0.95

0.46 0.78 0.95 0.95 0.95 0.49 0.70 0.95 0.95 0.95

0.44 0.72 0.95 0.95 0.95 0.46 0.64 0.95 0.95 0.95

0.40 0.66 0.81 0.95 0.95 0.42 0.57 0.79 0.95 0.95

0.39 0.59 0.76 0.81 0.94 0.41 0.51 0.73 0.80 0.89

0.38 0.56 0.70 0.77 0.81 0.40 0.49 0.68 0.76 0.80

< 4-6 >

1 -.0.8
I--__-

51E-01 I .E-01
1.1.,

o3

0.6

\
U

-,U----Um-.......

1.E-04 "1--1.E-06
lE-Ol 1 E -02

1 E-02

5.E-01 ,,-1 .E-01

0.4
* ' : u . . . .-

.... 11.-02 0.2 1.E-04 1,.E-06 0 0 I 0.2 I 0.4 Ohr 1 - 5-,E-01 " -u-1.E-01
-

I 0.6

I 0.8

0.8

--

.!-do--=1-

1.E-02
1.E-04 1.E-06
5.E-01

0.6

SA*

0.4 _ 0.2

1.E-01 -----1,E -0
-

.E-04 1.E-06

0 0

0.2

0.4 0/7r II " . .. ..

0.6

0.8

--in5.E-01 ... - 1.E -01 SI.E-02

0.8

....

A4 ...

0.6

I U-.

--

11E04 1.E-06
-

51-01 51E-01 .1E-02

0.4

0.2

1.E-04 I .E-06

0.2

0.4

0.6

0.8

Figure 4-3 Criticalcrack sizes at various failure probabilitiesfor example problem: random toughness and tensile properties, various loadings (Top: with 28-ksi [I 93-MPa] tension; center: with 8-ksi [55.2-MPa] tension and 20-ksi [137.9-MPa] bending; bottom: with 8-ksi [55.2-MPa] tension and 15-ksi [103.4-MPa] bending.) < 4-7 >

All of the inputs to the analyses whose results are shown in Figures 4-2 and 4-3 are considered to be well known. That is, there is no uncertainty due to a lack of data-all of the uncertainty is aleatoric and cannot be reduced by obtaining additional data [19]. The next section covers uncertainty resulting from limited data (epistemic uncertainty). There is yet another form of uncertainty, which is that due to taking a finite number of trials during the Monte Carlo simulation. Called computationaluncertainty, it can be reduced by taking more trials. There is computational uncertainty in the results shown in Figures 4-2 and 4-3, which can be characterized using commonly employed procedures, such as described on page 254 of Ang and Tang [20]. Defining the estimate of the failure probability, as F F=N where F is the number of failures and Nis the number of trials, the confidence interval of the true failure probability, p, is given by the expression Eq. 4-3

a!2 /2_

PO

+P

2N

N N

4N

Eq.4-4

ka/ 2 is related to the confidence interval and is equal to 1.65 for the 90% confidence interval. This equation is not applicable when/f = 0 or 1. As N goes to infinity, the confidence interval becomes narrower and --> p.

Equation 4-4 can be applied to the results of Figures 4-2 and 4-3 to estimate the computational uncertainty in these results. The case of 0/it = 0.7, fully saturated, 8 ksi (55.2 MPa) of tension, and 15 ksi (103.4 MPa) of bending is considered here. Table 4-3 summarizes the calculations, which start with the corresponding crack size results from Table 4-2.

<4-8 )1

Table 4-3 Summary of 90%confidence intervals for O/ir = 0.7, fully saturated, 8-ksi (55.2MPa) tension, and 15-ksi (103.4-MPa) bending

0.1 0.1
0.01

ld3X I U

U.4YYO

U.,UU4

0.77

5x10' 5x1 0'

0.0998 9.9XIcY, 9.3x1 O' 4.9x1 G'

2

0.1002
2 1.01x10"

0.70 0.56 0.38

10"

ilO'
Notes:

5x1 0
5

1.1x10,

2.Wx0'

1 From Table 4-2. 6 2 F=PN=(5x10)P

Figure 4-4 provides a plot of these results. The lines in Figure 4-4 are cubic spline fits to the bounds. It is seen that the confidence interval is very narrow, except for the 10-6 probability value.
0

-2

a..
O' -4

-6

-8 0 0.2 0.4 a/h 0.6 0.8

Figure 4-4 90% confidence intervals on probability for O/r = 0.7, fully saturated, 8-ksi (55.2-MPa) tension, and 15-ksi (103.4-MPa) bending The 90% confidence interval of a/h at the 10-6 probability is provided by the value of a/h at which the fits cross the 10-6 probability level. This is at a/h equals 0.340 and 0.407; the corresponding value from Table 4-2 is 0.38. Figure 4-5, which is Figure 4-3 with this interval indicated, shows that the 90% confidence level on the results is quite narrow. <4-9 >

1
0.8

"

...

"1-1tU

U.
0.6

,.-"-U"....in

1.E-02 1.E-04 1.E-06 -,-,,-- 1.E-01 1.E-02

0.4-5.E-01

0.2

Confidence interval.E-04 1.E-06

0 0

I 0.2

I 0.4

0.6

0.8

O/ln

Figure 4-5
Results of Figure4-3 with the 90% confidence interval shown for 0/ir = 0.7, 10' probobility (8-ksi [55.2-MPo] tension, 15-ksi [1 03.4-m1Pa] bending)

Correlated Toughness and Flow Strength with 4.3 Uncertainty

Results were generated that accounted for the correlation between the room temperature Cvat and high-temperature flow strength, with consideration given to the uncertainty due to the lack of data described in Appendix B. Only fully saturated conditions are considered. Computations were performed with 5x10' aleatoric trials and 50 epistemic trials. The 50 epistemic trials are sufficient to define the desired epistemic percentiles, and the 5x10' aleatoric trials are a tradeoff between computer execution time and the low probabilities to be estimated. Table 4-4 summarizes the results, which include epistemic quantiles of 0.1, 0.5, and 0.9. Monte Carlo simulations with additional aleatoric trials not included here showed that the 106 results are very close to the 10-5 results included in Table 4-4.

4 4-10 >

Table 4-4 Values of a/h for various 0/;r and selected probabilities consideringcorrelation and uncertainty (fully saturated)
Aleatoric Probability Epistemic Probability 0.1 0.29 0.31 0.33 0.49 0.63 0.80 0.95 0.95 0.95 0.28 0.29 0.30 0.44 0.55 0.72 0.95 0.95 0.95 0.26 0.28 0.29 0.42 0.50 0.67 0.95 0.95 0.95 0.25 0.26 0.28 0.39 0.46 0.59 0.79 0.8 0.83 0.25 0.26 0.27 0.38 0.44 0.55 0.73 0.75 0.77 0.25 0.26 0.27 0.37 0.43 0.51 0.68 0.7 0.72

10"5

0.5
0.9 0.1

0.01

0.5
0.9 0.1

0.5

0.5
0.9

0.1 10a 0.5 0.9 0.1

0.01

0.49 0.51 0.58 0.78 0.92 0.95 0.95 0.95 0.95

0.44 0.47 0.50 0.70 0.79 0.95 0.95 0.95 0.95

0.43 0.44 0.46 0.65 0.73 0.88 0.95 0.95 0.95

0.39 0.41 0.43 0.58 0.66 0.77 0.95 0.95 0.95

0.38 0.39 0.42 0.52 0.59 0.69 0.81 0.83 0.86

0.37 0.38 0.40 0.50 0.56 0.64 0.77 0.78 0.79

0.5 0.9 0.1

0.5

0.5 0.9

Figures 4-6 and 4-7 provide plots of the results in Table 4-4.

<4-11 >

0.8
-0.5

0.6
-- .

0.4 +

..- . . . . . .

0.01 10-5

. . . . . 0.2 +

8 ksi Tension, 20 ksi Bending 0 0 0.2

0.4

0 0.6

0 0.8

O/N Figure 4-6 Criticalcrack sizes at various failure probabilitiesconsidering correlation and uncertainty: 1Oth, 50th, and 90th epistemic percentiles (fully saturated,with 8 ksi [55.2 MPa] of tension and 20 ksi [137.9 MPa] of bending)
I

0.8 0.5 0.6 -t

I
0.4 -

..
-S

0.01
10 -5

-------

0.2 f 8 ksi Tension, 15 ksi Bending 0

!....

0.2

0 0.4 Ol/

0.6

0 0.8

Figure 4-7 Criticalcrack sizes at various failure probabilitiesconsideringcorrelation and uncertainty: 1Oth, 50th, and 90th epistemic percentiles (fully saturated,with 8 ksi [55.2 MPa] of tension and 15 ksi [103.4 MPa] of bending)
Runs were made that considered no uncertainty in the correlation (E = 0), and the results coincided with the median results in Table 4-4.

< 4-12 >

4.4 Comparison with Deterministic Analysis

In order to provide results comparable to the deterministic results shown in Figure 1-2, a probabilistic analysis was performed for a typical cold-leg-topressure-vessel joint. Table 4-5 shows the loads used. Table 4-5 Sample representativepiping interface loads

Load Case
Deadweight Operating basis earthquake Thermal

F. 1.3 193.8 13.7

F, 26.2 14.4 66.8

F. 0.6 87.3 6.0

M. 35.4 219.1 506.4

M, 20.3 8194.9 1595.0

M 2354.4 1141.7 7862.4

These loads result in the following moment and stresses: M = 15,028 in.-kips (1,698,014 N-m) and Ob = 10.28 ksi (70.9 MPa) (rounded to 10 ksi [68.9 MPa] in computer calculations). The maximum axial pressure stress is calculated to be
axial dR
_

2.25(16.0) 2(2.25)

8 ksi (55.2 MPa)

2h

for an assumed maximum design pressure, Pd, of 2250 psi (15.5 MPa). The PFM analysis was performed for CF-8M material in the fully aged condition using the uncorrelated material random variables. The results of the calculated critical flaw sizes for various failure probabilities are shown in Figure 4-8.

4 4-13 >

I
~.
%.

0.8 [
N.
-

0.6

~ --

5.E-01 .E-01

0.4

--

I.E-02 0.2
--

I..E-04

1.E-06

0~ 0

i 0.2

I 0.4

0.6

0.8

elf
Figure 4-8 Criticalcrack sizes at various failure probabilitiesfor example problem with 8-ksi (55.2-MPa)tension and 1O-ksi (68.9-MPa) bending stress for fully aged CF-8M
Figure 4-9 provides a plot of the results of Figure 4-8 along with the deterministic results shown in Figure 1-2. It is seen that the deterministic results with a safety factor of one correspond closely to the probabilistic results for 10-4 failure probability, and the deterministic results with the higher safety factors fall well below the 10.6 results. If 10-6 is a reasonable target failure probability, this difference dearly illustrates the degree of conservatism in the bounding deterministic analysis. As previously indicated, the acceptance criteria are not part of the scope of this study.

< 4-14 >

EPFM Solution for o,, = 8 ksi and ob = 10 ksi

I0.8-

1 0.4-

0 0

1 0.2 0.4 0.6 0.8

Length of Flaw, Fraction of Circumference (Ubicr) *20% DFN (SFm = SFb = 1.0) N 20% DFN (SFm = 2.7, SFb = 2.3)

0 30% DFN (SFm

A,30% DFN (SFm = SFb =1.0) = 2.7, SFb = 2.3)

DFN

= delta

ferrite number

Figure 4-9 Probabilistic results of Figure 4-7 plotted with corresponding deterministic results from Figure 1-2

4.5

Results with Delta Ferrite as Input

Results presented in the preceding parts of Section 4 considered the degradation parameter (D as the input random variable. As covered in Section 1, the delta ferrite content is taken as the governing variable when identifying piping components of concern (see Figure 1-1). Therefore, the analyses were also made with a specified value of the ferrite content. Figure 4-10 provides results for various specified delta ferrite contents. The composition was held constant at the alloy content given at the beginning of Section 4.1. Runs were made using the delta ferrite distribution of Figure A-23 split at >20% and <20%. Additional runs were made with normally distributed delta ferrite with the means and standard deviations indicated in Figure 4-10.

<4-15 >

1.0

0.8

0.6

0.4

0.2

0 ---'-

>20 <20 all

25,2

-0.0
I I I

9-

22,6

0.2

0.4

0.6

0.8

a/h Figure 4-10 C6 probobilityfor vorious delto ferrite contents, 8-ksi (55.2Criticol crock sizes ot 1
MPo) tension, ond 15-ksi (1 03.4-MPa) bending

Figure 4-10 shows that the delta ferrite content does not have a strong influence on the critical crack sizes within the range considered. 4.6 Tabular Presentation for Probability of Failure of 10' and 10' Tables 4-6 and 4-7 provide values of critical crack depths at the 10-6 and 10-4 probability levels for the 32-in. (812.8-mm) cold leg pipe of the example problem. The results are presented for various stress ratios, where the stress ratio is defined as
+ Crb Stress Ratio = onm aflo

<4-16>

Table 4-6 Criticalcrack sizes for 32-in. (812.8-mm) line at the 10' probability level

U.OU

U. 13UU

U. 14UU

U. 14UU

U. IJ UU

U. IJUU

U. 14VV

0.55
0.50

0.2300
0.3000

0.2100
0.2900

0.2100
0.2800

0.2000
0.2600

0.2000
0.2600

0.2000
0.2600

0.45
0.40

0.4100
0.5600

0.3700
0.4900

0.3600
0.4600

0.3400
0.4300

0.3300
0.4100

0.3300
0.4000

0.35
0.30

0.7500
0.9500

0.6700
0.8100

0.6100
0.7600

0.5500
0.6800

0.4900
0.6100

0.4800
0.5800

0.25
0.20

0.9500
0.9500

0.9500
0.9500

0.9500
0.9500

0.7800
0.8900

0.7100
0.7800

0.6600
0.7400

0.15

0.9500

0.9500

0.9500

0.9500

0.9300

0.8100

Table 4-7 Criticalcrack sizes for 32-in. (812.8-mm) line at the 10' probabilitylevel

0.60

0J..00

0.J30U

0.3100

0. 3000

0.2900

0.2900

0.55 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15

0.4600
0.6000

0.4000
0.5100

0.3900
0.4800

0.3700
0.4400

0.3500
0.4200

0.3500
0.4200

0.7500
0.8900

0.6700
0.7800

0.6100
0.7300

0.5500
0.6600

0.4900
0.5900

0.4800
0.5700

0.9500
0.9500

0.9500
0.9500

0.8600
0.9500

0.7600
0.8200

0.6800
0.7600

0.6400
0.7100

0.9500 0.9500 0.9500

0.9500 0.9500 0.9500

0.9500 0.9500 0.9500

0.9500
0.9500

0.8200
0.9500

0.7700
0.8200

0.9500

0.9500

0.9500

The flow stress ofi. is taken to be 57.1 ksi (393.7 MPa) in this example. The membrane stress, oa., is held constant at 8 ksi (55.2 MPa), which corresponds to an internal pressure of 2.25 ksi (15.5 MPa). The bending stress, ab, is varied to produce the desired stress ratio in Tables 4-6 and 4-7. In total, 10' trials were used, with random D as the input. Tables 4-8 and 4-9 provide similar results for 12.75-in. (323.9-mm) (outer diameter) line with a thickness of 1.312 in. (33.3 mm). The pressure is again held constant at 2.25 ksi (15.5 MPa), which, for this size of line, corresponds to a membrane stress of 4 ksi (27.6 MPa).

< 4-17>

Table 4-8
Criticalcrack sizes for 12.75-in. (323.9-mm) line at the 10 probability level

0.60

0.2100

0.1900

0.1800

0.1700

0.1700

0.1800

0.55 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15

0.2800 0.3800 0.5100 0.7300 0.9500 0.9500 0.9500 0.9500 0.9500

0.2700 0.3600 0.4600 0.6600 0.8000 0.9500 0.9500 0.9500 0.9500

0.2600 0.3300 0.4400 0.5900 0.7600 0.9500 0.9500 0.9500 0.9500

0.2400 0.3200 0.4100 0.5200 0.6700 0.7900 0.9500 0.9500 0.9500

0.2400 0.3100 0.3900 0.4800 0.6000 0.7200 0.8000 0.9500 0.9500

0.2400 0.3100 0.3800 0.4700 0.5700 0.6700 0.7600 0.8200 0.9500

Table 4-9 Criticalcrack sizes for 12.75-in. (323.9-mm) line at the I W probability level

0.60 0.55 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15

0.4200 0.5400 0.7200 0.8400 0.9500 0.9500 0.9500 0.9500 0.9500 0.9500

0.3800 0.4800 0.6300 0.7600 0.9500 0.9500 0.9500 0.9500 0.9500 0.9500

0.3700 0.4600 0.5700 0.7000 0.8300 0.9500 0.9500 0.9500 0.9500 0.9500

0.3500 0.4200 0.5100 0.6400 0.7600 0.8200 0.9500 0.9500 0.9500 0.9500

0.3300 0.4000 0.4700 0.5800 0.6700 0.7600 0.8200 0.9500 0.9500 0.9500

0.3300 0.4000 0.4600 0.5600 0.6300 0.7100 0.7700 0.8200 0.9500 0.9500

(4-18 >

Section 5: Discussion
The PFM model developed in the preceding sections can be used to perform flaw tolerance evaluation of embrittled CASS piping to determine the allowable flaw sizes at a given probability of failure. In this study, grade CF-8M material was considered because it is the most susceptible to thermal aging. The model can be used for other grades of CASS using the NUREG databases of material properties (strength and toughness) for the other CASS alloys. Limitations on the use of the PFM model to Grade CF-8M are provided in Section 1.3. The PFM model was exercised for a number of sample cases to demonstrate the sensitivity of the results for three probabilistic conditions and various assumed tension and bending loading conditions. The effect of random toughness was theprimary concern for these analyses because the variation of strength and toughness in the aged condition is one of the greatest uncertainties for the CASS materials. For CF-8M materials with high delta ferrite content, a bounding deterministic analysis produces very conservative maximum allowable flaw depths that could overly penalize the CASS piping materials when evaluating flaw tolerance. This is shown in Figure 1-2. The series of sample cases in Figures 4-2, 4-3, and 4-6 through 4-9 shows the benefits of or improvements from not having to assume the worst case (that is, the lower bound) toughness. Figure 4-2 shows the effect of the distribution of saturated material toughness with known chemical compositions. It can be seen that, for both aged and unaged materials, the minimum critical flaw depth is above or close to 40% of the total wall thickness, even with very long circumferential flaw length (up to 70% of the total circumference). Also, the difference in critical flaw depth between the 0.5 quantile and 10"6 th quantile is not significant. The smallest difference between the 0.5-quantile and 10 6th-quantile lines is as small as 5% of the total wall thickness. The thermal aging effect based on Figure 4-2 is not critical. The difference is 2-9% of the total wall thickness, smaller than the difference between the 0.5-quantile and 10-6th-quantile lines. Figure 4-3 presents the critical crack depths for aged and unaged materials for various loads, randomizing toughness, tensile properties, and alloy content. No correlations are considered in Figure 4-3. Figures 4-6 and 4-7 present results considering correlations between toughness and tensile properties, with uncertainty bands representing the epistemic uncertainty due to limited data. A direct comparison of correlated and uncorrelated results at low quantiles is not possible with current results. The correlated runs were made with several trials <5-1 >

that did not allow determination of the 10-6 probability because of the number of epistemic trials run (50). However, for the case of 8-ksi (55.2-MPa) tension and 15-ksi (103.4-MPa) bending, a comparison of correlated 10' results with uncorrelated 10-6 results shows these two cases to be very nearly identical. (Figure 4-5 shows that the epistemic spread in the 10-1 results is small.) Because the correlated 10- result closely corresponds to the 106 result, the correlated results are less favorable, but the 10-5 and 106 results for the correlated case are very nearly the same. (The same crack size that gives 106 failure probability for uncorrelated gives 10' for correlated.) As shown in Figures 4-6 and 4-7, the spread in the results at low probabilities due to epistemic uncertainty is small, indicating that more data to reduce uncertainty are not needed. However, the treatment of uncertainty in Appendix B is subject to change if more data become available and a more rigorous treatment is identified. As observed in Figure 4-3, for the same percentile, the aged critical flaw sizes are often larger than the unaged critical flaw sizes, which is contrary to expectations for a thermally aged material with decreased toughness. However, it should be noted that with the decreasing of toughness, the tensile properties increase. The median values of yield strength increase from 26.47 ksi (182.5 MPa) to 38.51 ksi (265.5 MPa) (see Appendix A). Considering that the applied stress (tension plus bending) is 28 ksi (193.1 MPa) in the upper two plots of Figure 4-3, plasticity would be significant, and the change of yield strength might largely change the applied J-T solution because of applied stresses. (A higher yield strength would reduce the plastic component ofJ as well as the effect of the plasticity correction to the elastic component ofJ.) This effect would counter the effect of decreasing toughness. Therefore, it is important to consider the thermal aging effect on both toughness and tensile properties for the evaluation of thermal aging of CASS. However, the improvement in flaw tolerance for aged materials is not uniformly apparent for all stress levels. For example, Figure 4-3 shows that the aged material is better at high stress levels (the dashed lines at higher a/h at the top of Figure 4-3). At lower stress levels, the trend is not so consistent. This is because the plastic component ofJ 0 ,ppl,~has a much greater dependence on the yield strength of the material. The advantage of using a PFM analysis to determine the allowable flaw sizes for embrittled CASS piping components can be seen in Figure 4-9. By using a deterministic approach with lower-bound material properties and ASME Section XI Code structural factors, the allowable flaw depths are less than10% of wall thickness, which would have rendered this particular pipe under the loading condition almost impossible to operate. However, with a PFM analysis and accepting the probability of failure of 106 as a target, the allowable flaw depths are relatively large. The relatively large allowable flaw depths also make it easier for the flaw to be detected with the ever-improving UT inspection techniques for CASS piping components. Tables similar to Tables 4-6 and 4-7 can be

<5-2 >

developed on a component-specific basis once the stresses are known and the material database is established. When combined with recent developments in nondestructive evaluation, the methodology presented in this report can be used as part of the development of a Code Case for management of CASS piping components. This is not part of the scope of this study.

<5-3>

Section 6: Summary
A PFM methodology for determining the flaw tolerance of thermally aged CASS piping components has been presented. The need for such methodology stems from the fact that traditional deterministic fracture mechanics methods are conservative and could result in relatively small acceptable flaw sizes, which could render management of these components very difficult in license renewal space. The methodology was developed using Grade CF-SM because it is the most susceptible to thermal aging, with the limitations stated in Section 1.3. The PFM model is based on the extensive data gathered in several NUREG reports, for example, NUREG/CR-4513 [17]. Crack stability is treated by aJintegral-based tearing instability. Random variables include material strength and the crack growth resistance curve (J-da)as well as delta ferrite and alloy content. Applied loads are treated as deterministic. The output of the probabilistic model is crack sizes (depth and circumferential extent) that would become unstable (fail) with a given probability when specified loads are applied. The PFM model predicts that crack sizes for the saturated material and unaged material are similar, with the aged material in some cases being more flaw-tolerant. (There are competing effects between lower toughness and higher strength of the aged material.) At the 10-6 probability of failure level, critical crack depths vary from 10% to 50% of the wall thickness, depending on the applied loads. The crack depths increase to 30-50% for loadings expected during normal operating conditions in hot leg primary piping in PWRs. This is contrasted with a critical crack depth of about 10% wall thickness obtained when using a conservative deterministic analysis with conservative margins and comparable loading. Correlations between toughness and flow strength were explored, but they suffered from a lack of data. Preliminary treatment of the effects of a lack of data suggest that uncertainty due to lack of data does not have a large influence on the results. The relatively large critical flaw sizes predicted with a 106 failure probability when normal service loads are applied suggest that CASS piping is quite flawtolerant, even in the embrittled condition; these results should be useful in developing ASME Code flaw acceptance standards for high-delta-ferrite materials.

<6-1 >

The focus of the study covered in this report was the development of the flaw tolerance methodology and not the acceptance criteria. The justification for an acceptable failure probability will be addressed outside of this study. The study described in this report can be used in part to develop an ASME Section X0 Code Case for long-term management of CASS piping components.

< 6-2>

Section 7: References
1. Generic Aging Lessons Learned report, NUREG-1801, Rev. 2, December 2010. 2. 3. Nondestructive Evaluation:Flaw Tolerance Evaluationof Thermally Aged Cast Austenitic Stainless SteelPiping.EPRI, Palo Alto, CA: 2009. 1019128. Assessment of Crack Detection in Heavy- Walled Cast Stainless Steel Piping Using Low-Frequency UltrasonicMethods. U.S. Nuclear Regulatory Commission report NUREG/CR-6933, 2007.

4. An Evaluation of UltrasonicPhasedArray Testingfor CastAustenitic Stainless Steel PressurizerSurge Line Weld. U.S. Nuclear Regulatory Commission report NUREG/CR-7122, 2012. 5. T. Griesbach, V. Marthandam, H. Qian, and P. O'Regan, "Flaw Tolerance Evaluation of CASS Piping Materials." ProceedingsofPVP2009ASME Pressure Vessels and Piping Division Conference (Prague, Czech Republic, July 26-30, 2009). PVP2009-77421, 2009. W. Bamford and E. Landerman, "Thermal Aging of Cast Stainless Steel and Its Impact on Piping Integrity." JournalofEngineeringMaterialsand Technology, Volume 107, January 1985. C. Buchalet, Y. Meyzaud, and P. Taupin, "Effect of Long-Term Aging on the Critical Defect Sizes in Primary Piping Cast Stainless Steel Elbows." Proceedingsof the Second International Symposium on Environmental Degradation ofMaterials in Nuclear Power Systems- Water Reactors (Monterey, California, September 9-12, 1985). A. Diaz, S. Doctor, B. Hildebrand, F. Simonen, G. Schuster, E. Andersen, G. McDonald, and R. Hasse, Evaluationof UltrasonicInspection Techniques for Coarse-Grained Materials.NUREG/CR-6594, PNNL-11171, Pacific Northwest National Laboratory, October 1998. S. Lee, P. Kuo, K. Wickman, and 0. Chopra, "Flaw Evaluation of Thermally Aged Cast Stainless Steel in Light-Water Reactor Applications." International Journalof Pressure Vessels and Piping,72 (1997), pages 37-44.

6.

7.

8.

9.

10. D. Gavenda, W. Michaud, T. Galvin, W. Burke, and 0. Chopra, Effects of ThermalAging on Fracture Toughness and Cbhapy-Impact Strength of Stainless Steel Pipe Welds. NUREG/CR-6428, 1995. 11. An EngineeringApproach for Elastic-PlasticFracture Analysis. EPRI, Palo Alto, CA: 1981. NP-1931.

< 7-1 >

12. T. Anderson, FractureMechanics."FundamentalsandApplications,Second Edition. CRC Press, 1994. 13. Ductile FractureHandbook: Volume 2. EPRI, Palo Alto, CA: 1989. NP-6301DV2. 14. D. Cho, H. Seo, Y. Kim, Y. Chang, M. Jhung, and Y. Choi, "Advances in JIntegral Estimation of Circumferentially Surface Cracked Pipes." Fatigueand FractureofEngineeringMaterialsand Structures, Volume 34, Issue 9 (September 2011), pages 667-681. 15. Y. Kim, J. Kim, Y. Lee, and Y. Kim, "Nonlinear Fracture Mechanics Analyses of Part Circumferential Surface Cracked Pipes." International Journalof Fracture,Vol. 16 (2002), pp. 347-375. 16. D. Harris, E. Lim, and D. Dedhia, Probabilityof Pipe Fracturein the Primary Coolant Loop ofa PWR Plant, Vol 5: ProbabilisticFractureMechanicsAnalysis. U.S. Nuclear Regulatory Commission report NUREG/CR-2189, Vol. 5, Washington, D.C., 1981. 17. 0. Chopra, Estimation ofFractureToughness ofCast Stainless Steels During ThermalAging in L WR Systems. U.S. Nuclear Regulatory Commission report NUREG/CR-4513, Rev. 1, 1994. 18. W. Michaud, P. Toben, W. Soppet, and 0. Chopra, Tensile Property of Thermally Aged Cast Stainless Steels. U.S. Nuclear Characterization Regulatory Commission report NUREG/CR-6142, 1994. 19. A. Der Kiureghian and 0. Ditlevsen, "Aleatoric or Epistemic? Does it Matter?" StructuralSafety, Vol. 31 (2009), pp. 105-112. 20. A. Ang and W. Tang, ProbabilityConcepts in EngineeringPlanningand John Wiley & Sons, 1975. Design, Volume 1: Basic Principles. 21. 0. Chopra and A. Sather, InitialAssessmentofthe Mechanisms and Significance of Cast Stainless Steels in LWR Systems. U.S. Nuclear Regulatory Commission report NUREG/CR-5385, 1990. 22. 0. Chopra, Long-Term Embrittlement ofCast Duplex Steels in L WR Systems. U.S. Nuclear Regulatory Commission report NUREG/CR-2744, Vol. 7, No.1, 1993. 23. 0. Chopra, Long- Term Embrittlement ofCast Duplex Steels in L WR Systems. U.S. Nuclear Regulatory Commission report NUREG/CR-2744, Vol. 6, No. 1, 1992. 24. 0. Chopra, A. Sather, and L. Bush, Long-Term Embrittlement of CastDuplex Steels in L WR Systems. U.S. Nuclear Regulatory Commission report NUREG/CR-2744, Vol. 4, No. 2, 1991. 25. W. Mendenhall et al., Probabilityand Statistics, Thompson, 2003. 26. S. Weinberg, Applied LinearRegression, Wiley-Interscience, 2005.

< 7-2 >

Appendix A: Characterization of MaterialRelated Random Variables

The material properties that enter into the evaluation of critical crack sizes vary with exposure time, temperature, and material composition. There is considerable scatter in these properties, and the purpose of this appendix is to describe the statistical distributions of the relevant properties. From the initial unaged condition, the CASS material properties gradually change until reaching the fully aged or saturated condition, after which there are no further changes with continued exposure. This appendix considers the unaged and saturated conditions, and it does not consider correlations between random variables. Appendix C considers possible correlations between the fully saturated room temperature Charpy energy and the fully saturated high-temperature flow strength. CF-8M material is the focus because, among the cast austenitic materials employed in commercial nuclear plants, this is the alloy that is most degraded by thermal aging. A. 1 Unaged Properties

The unaged material properties of CF-8M at elevated temperature are characterized in Sections A.1.1 and A.1.2. A. 1. 1 Unaged Tensile Data on the unaged tensile properties of CF-8M at 650'F (343C) were gathered from Table B-4 of Nondestructive Evaluation:Flaw ToleranceEvaluation of Thermally Aged CastAustenitic Stainless Steel Piping [2]. This provided 24 values of the yield and ultimate strengths, which then provided 24 values of the flow strength (which is the average of the yield and ultimate strengths). The mean and standard deviations as calculated from the data are summarized in Table A-1.

<A-I >

Table A- I Values of the mean and standard deviation of the tensile properties of unaged CF8M at 6507F (343QC)
Yield strength, ksi 26.471 2.917

Ultimate strength, ksi

Flow strength, ksi

79.200
52.838

7.864
5.202

Figures A-1 through A-3 provide normal probability plots of the data along with the corresponding fits using the means and standard deviations from Table A-1.
0.98 0.95 0.9 0.8 " 0.7

0.6 IL 0.5

0.4
S0_3

2D

22

24

26

28 30 YIed Strength, kW

32

34

Figure A- I Data and fitted normal distribution of yield strength for unaged CF-8M at 650F (3430C)

( A-2 >

na OA

0.95

0 0 0 0 0

0.9 1 0.8
0.7

0,6 0.5
0.4 0.3 0.2

2 2
0 0 0 0 0

E
00

0.1 0.05

60

65

70

75

80

85

90

95

100

Ultimate Strength, ksl

FigureA-2 Data and fitted normal distribution of ultimate strength for unaged CF-8M at 6507F (343C)
0,995 0.99 0.98 0.95 0.9 0.8 0.7 a 0.6 0.5 0.3 02
I I I I I I I

_ 0.4
0 S /

E o

0.1 0.05 0.02

0.01

1
30
35 40

0~
| | |

45

50 Flow Strength, ksl

55

60

65

70

Figure A-3 Data and fitted normal distribution of flow strength for unaged CF-8M at 650F (343C)

< A-3 >

The distributions of the unaged yield, flow, and ultimate strengths have now been defined. A description of the entire stress-strain relationship is needed for evaluation of the plastic component of the applied value of the J-integral. The conventional Ramberg-Osgood relationship is employed, and the procedure outlined in NUREG/CR-6142 [18] is followed. The following form of the RambergOsgood relationship is used:

E -a
0O'flo

0
fo]

Eq. A- I

ao is the average of the yield and ultimate strength values. Young's modulus is taken to be 25,600 ksi (176,506 MPa). Equation A-1 can also be written as
flO

C =-+ E

E (.o-] 20)

a-

Eq. A-2

Following NUREG/CR-6142 [18] closely, it is stated on page 52 that "the parameter n does not change with thermal aging" and that the parameter 0 is related to the flow strength by the relationship = a, +b3a-flo Eq. A-3

Concentrating on strains less than 5% (because the nominal applied strain levels experienced by pipes in service are generally less than 5%), Table 8 on page 53 of NUREG/CR-6142 [18] gives the following values for CF-8M at 290C (554'F): " * *
a3 =

145.9

b3 =-0.314 n = 6.6

The preceding values are for stresses in MPa. Figure A-4, which is Figure 24 of NUREG/CR-6142 [18], shows the data upon which the preceding values of a3 and b& are based.

< A-4 >'

100 100 -1 I..

,...~.......,,...

CF-8M Steel 29000

0 00
500

0 0 0

0
a, = 145.9 - 0.314cy,

00
008 0'
1
350 Flow Stress, G, (MPs) 400 450

0
250

' ' ' 'I 1 I 1 1 1 1

300

Figure A-A Data and relationshipbetween a and flow stress for CF-8M [A2] A great deal of scatter is observed in the a-Oyo, relationship, and a negative value of a is predicted for stresses above 456 MPa (66.1 ksi). Negative a results in negative plastic strain for positive stress, which is unrealistic and could occur during Monte Carlo simulation with many trials. Following NUREG/CR-6142, the value of n = 6.6 and Equation A-4 are applicable to both unaged and fully aged materials (as well as intermediate aging). The value of n = 6.6 will be used, along with a being a function of flow stress. However, a will be re-evaluated, and a different functional form between a and the flow stress will be employed. Attention is focused on the stress-strain results for unaged and saturated material. NUREG/CR-6142 [18] contains tabulated data. Table A-2 summarizes the available data for CF-8M, with four unaged and six saturated specimens available.

4 A-5 >

Table A-2 Summary of 290'C (5540F) stress-strain data available for CF-8M unaged and saturated [A2]
Heat 74

75

74

75

Data from Table 5 o NUREG/CR-6142 [18]

Specimen number Table number 0r, MPa q, MPO 744-40 294 171.8 412.3 297.9 43.21 86.188 9.360 743-42 95 164.6 442.8 297.9 43.21 60.021 7.722 754-40 350 196.8 470.8 333.5 48.37 25.078 7.034 753-42 351 191.5 474.8 333.5 48.37 44.831 8.314 742-28 304 203.5 473.3 353.3 51.24 37.451 5.691 742-29 305 215.2 474.4 353.3 51.24 40.247 7.600 74-130 306 235.4 517.7 353.3 51.24 36.297 8.371 752-28 360 258.9 587.9 415.0 60.19 19.436 7.167 752-29 75-130

361 264.1 581.7 415.0 60.19 16.035 6.185 361 19 6.6 16.62

362 239.6 557.8 415.0 60.19 27.785 6.506 362 16 6.6 26.74

a1, MPa
cg,, ksi A N

Tabulated Stress-Strain Data [A2]

Table number
Data

294 26

95 22

350 14

351 16

304 12

305 30

306 24

360 19

Values Derived from Tabulated Stress-Strain

N A 6.6 43.26 6.6 46.60 6.6 30.19 6.6 35.02 6.6 46.38 6.6 35.54 6.6 34.44 6.6 17.33

The stress-strain data are tabulated in NUREG/CR-6142 [18] for each of the specimens in Table A-2. Figure A-5 provides plots.

< A-6 >

80

60

400

20

0.01

0.02 Strain

0.03

0.04

0.05

Figure A-5 Stress-strain data for unaged and saturatedCF-8M, 2900 C (5547tF) (The open points and dashed lines are unaged; others are saturated.) The upper part of Table A-2 shows the values of a and n for each of the specimens, and considerable variation from specimen to specimen in both a and n is observed. In order to reduce the scatter shown in Figure A-4, the value of n was fixed at 6.6 and a value of a obtained by least squares for each of the 10 specimens in Table A-2. This provided the values of a shown at the bottom of Table A-2. A curve fit to the a - aop data was then performed using a functional form that will not provide a negative a at higher stresses. Figure A-6 is a plot of the data and the curve fit.

< A-7 >

50

40

30

20

10

0
40

50

60

70

80

Flow Stres, kli

Figure A-6

The a - a, data and fit for unaged and fully aged CF-8M
The following relationship between a and afo, is shown in Figure A-6:

a=

c1a";o -- C 3

Eq. A-4

where cl = 0.008848,

c2

= 0.64778, and

c3

= 0.08077.

These values of the constants are for stresses in ksi, and they result in a negative a if c 3 > cl o , which corresponds to aflo = 30.1 ksi (207.5 MPa). This low value can occur with a low probability, and if it does appear during a Monte Carlo simulation, the sampled flow stress is set equal to 32 ksi (220.6 MPa). Figure A-7 is a plot of the fitted curves and the corresponding stress-strain data for each of the 10 specimens. Good fits to the data are observed.

4 A-8 )

60

* 8 T * 0 0 A
-

T294 T295 T305 T305 T350 T361 T360

T361

-.

0-V

-T294 --T235 T304

T305

. . . ..... .T30
- - T360

____T361

T362

0.01

0.02 Strain

0.03

0.04

0.05

Figure A-7 Comparison of data and fits for each of 10 specimens included in Table A-2 Summarizing, the following procedure is used to define a random stress-strain curve for unaged material: 1. n is fixed at 6.6. 2. aflis sampled from the normal distribution of unaged flow stress, shown in Figure A-3. 3. a is calculated from Equation A-4.

4. Equation A-2 gives the stress-strain relationship. A Monte Carlo simulation with 1000 trials was performed to generate 1000 stress-strain curves for unaged material. Various quantiles of the curves were obtained and are compared in Figure A-8 with the data for the unaged specimens. A value of E = 25,600 ksi (176,506 MPa) was employed.

A A-9 >

80

~7Q

90 . . ... ... .

0.01

0.02

0.03

0.04

0.05

Strain

Figure A-8 Data and quantiles of CF-8M unaged tensile tests included in Table A-2 At small strains, the data in Figure A-8 fall dose to the median, but at larger strains (>-0.005), the data fall around the predicted 10th percentile. A. 1.2 Unaged Toughness A statistical characterization of the crack growth resistance curve is needed for the probabilistic analysis of the reliability of CASS piping. Efforts focus on CF8M material. The correlations of Chopra as summarized in NUREG/CR-4513 [17] are the starting point in these efforts. These correlations depend heavily on the Charpy impact energy, Cv. Following NUREG/CRs 4513, 5385, and three volumes of 2744 [17, 21-24], the crack growth resistance curve is expressed as j = CR (Aa) m Eq. A-5

The Chopra correlations for defining the crack growth resistance curve are based on the Charpy energy. A. 1.2.1 Impact Energy (Unaged, Room Temperature)

Data on the unaged Charpy energy at room temperature were obtained from Appendix B of NUREG/CR-5385 [21]. The data are summarized in Table A-3.

< A-10 >

Table A-3 Summary of unaged charpy data for CF-8M at room temperature [21]
62 63 64 65 66 P4 70 74 75 B-14 B-14 B-16 B-18 B-19 B-27 B-36 B-38 B-40 195.5, 243.1, 234.0, 239.3 229.6, 233.9, 293.2, 221.9, 245.4 62.6, 165.5, 225.4, 220.2, 187.7 199.1, 238.5, 226.1, 222.5 168.0, 259.6, 202.3, 251.9 228.2, 226.4, 190.3, 265.8 395.0, 325.2 267.9, 157.0, 205.6 260.6, 247.5, 237.6, 201.8

The data of Table A-3 were converted to ft-lb by multiplying by 0.59 (see page B-1 of NUREG/CR-5385), then the average and standard deviation of the logarithm of the combined data set were evaluated. The data were also sorted and plotted as a cumulative probability distribution. Figure A-9 is a plot of the data on lognormal scales, along with the corresponding fit using the mean and standard deviation of the logarithms. The agreement between the data and the fit is not that good, but a lognormal distribution will be used. The lognormal probability density function will be written as

p(Ccv) -lIiv 1

2, 2

Eq. A-6

The values of the median and shape parameter of the unaged room temperature data were Cvso = 130.3 ft-lb (176.7 J) and = 0.279. The unaged Charpy energy distribution will be represented by a lognormal distribution.
14

<A-1 I>

CF-SM Unaged Room Temperature 0.99 0.98 data 0.95 i

0.9 g0.8 0.7 0.6

0.

fit

I
U

0.5 0.4 0.3 0.2 0.1 0.05 0.02

I I I I I I

E
C.)

50

102

2x10

Cv, ft-lb

Figure A-9 Cumulative distribution of unaged CF-8M Charpy at room temperature (data from Table B of NUREG/CR-5385)

A.1.2.2

Resistance Curve Coefficient (Unaged, High Temperature)

Once the Charpy energy is defined, the coefficient in the crack growth resistance curve (see CR of Equation A-5) is obtained from the correlation shown in Figure A-10, which is Figure 11 of NUREG/CR-4513 [17]. Figure A-10 is applicable to aged and unaged material.

( A-12 >

RT Impact Energy (ft-lb) 5

-~ I

10
I I

50
I I
1111111

100
11111 I 4-

II

liii I

290 - 320'C 1000C =102Cv0,

C 0

UJ

100-

""

C =102C `

=49Cv'

[.i

10

EPRI

Open Synbolos: CF-3. CF-8 Closed Symbolds: CF-8M 10

-4 i 1 i i 1i i i I ! I ii I i

ANL
i 500

10

50

100

(J/cm') RIT kripact Energy, C~,

Figure A- 10 Figure 11 from NUREG/CR-4513 showing data points for high-temperatureC, vs. room temperature Charpy impact energy The value of the exponent m in Equation A-5 is defined knowing the value of CR by the relationship shown in Figure A-11. Note that in Figure A-11, n is used instead of m.
1
I i

CF-8M Steel
290C

i
Ii

0 0 0 0 00
t I 1 i I J

0 0 00 0 0

cx o
0 0

n = 0.23 + 0.06log,l(C,)
! I!

0 10

100 Charpy Impact Energy C. (J/cm')

Figure A- Il Figure 12 of NUREG/CR-4513 [17] showing relationshipbetween exponent n and Charpy impact energy ( A-13 >

Following this approach, the following relationship for CF-8M from Figure A10 will be used to describe the value of CR for a given Charpy energy for all aging conditions:
CR = 57C1 1
04

Eq. A-7

The coefficient 57 is for Charpy andJ-R in metric units. The randomness in the data will be considered to be in the Charpy data, but for a given Charpy value, the coefficient of the crack growth resistance curve is deterministically defined by Equation A-7. The following are Equations 3.2.25a and 3.2.26a from NUREG/CR-4513 [17], which are applicable to CF-8M at 5500F (288'C) and provideJin in-lb/in 2 when Aa is in inches. J = {404(25.4)"m(Cvsat )041 [Aa]"' m = 0.244 + 0.06 log,0 (CIa,) Eq. A-8 Eq. A-9

Comparing Equations A-5 and A-8, the value of CR can be expressed as

CR
=

404(25.4")C0"4'

Eq. A-10

The value of Cvs,, to be used in Equations A-7 and A-8 is the room temperature value, and theJ-R curve is for 550'F (288'C). Equations A-8 through A-10 are actually applicable to any Charpy value, not just the saturated value, as can be seen in Figure A-10. Therefore, these equations can be written more generally as J = {404(25.4)m (C, )0.41 JAa]" m = 0.244+ 0.06log ,0 (C,)
CR = 404(25.4" )C.
4

Eq. A-8a Eq. A-9o Eq. A- 1Oa

NUREG/CR-4513 [17] has lower-bound relationships for the J-R curve. For CF-8M with more than 15% ferrite, Figures 3 and 4 of the NUREG provide the following expressions (for in-lb/in 2 and inch) J = 3035Aa"3 " J = 2599Aa 3 centrifugally cast statically cast Eq. A- I la Eq. A-1 Ib

4 A-14 >

It was hoped that the data in Figures A-10 and A-11 could be used to obtain values of CR and m for each specimen, thereby allowing the correlation between these two values to be analyzed. However, the data of Figures A-10 and A-11 could not be used to define specimen-by-specimen values of CR and m.

A.2

Saturated Properties

Properties for the fully saturated CF-8M material are characterized in Sections A.2.1 and A.2.2.

A.2.1 Saturated Tensile, High Temperature

A statistical characterization of the distribution of the fully aged stress-strain curve is needed for the probabilistic analysis of the reliability of CASS piping. These efforts are concentrated on CF-SM material. The stress-strain properties are needed for computation of the plastic component of the applied value of the J-integral that, in turn, is used with the crack growth resistance relationship to assess crack instability. The tensile properties change with time and temperature. Figure A-12 shows the flow stress ratio data from NUREG/CR-6142 [18] for CF-8M steel at 290C (5530F). A great deal of scatter is observed. (P and 0 depend on composition, temperature, and time. The time variation is logarithmic.)
1.4
-' 4.

CF-OM Steel 290C

ig
0N
C

00o
00 0

0.
0

0.00

0 8

~12 0ai

U.

Ratio = 0.69 + 0.14(P-0+2.9) 1.00 % Ratio !c 1.24

0.8 0 1 2 3 Normalized Aging Parameter (P 4 5

+ 2.9)

Figure A- 12 Reproduction of Figure 21, page 50 of NUREG/CR-6142 [A2]

<A-15 >

The line shown in Figure A-12 is the mean minus one standard deviation fit suggested in NUREG/CR-6142 [18]. The aging saturates as time increases, eventually reaching a plateau value of the aged flow stress divided by the unaged flow stress. Taking the data points for the normalizing aging parameter greater than or equal to 3 as being representative of the plateau, there are 14 data points. Figure A-13 is a normal probability plot of these 14 data points along with the line corresponding to the mean and standard deviation of the data mean = 1.189 and standard deviation = 0.071.
0.98 0.95 0.9 0.8 0.7 0.6

I
E
U

0.5 0.4 0.3 0.2 0.1 0.05 0.02

1 1.05 1.1 1.15 1.2 1,25 1.3 1.35 1.4

Ratio

Figure A- 13 Normal probability plot of the flow stress ratio for the data of Figure A- 12 with the normalizing parametergreater than 3 (Mean = 1. 189; standarddeviation = 0.071.) A good representation of the data is observed. The preceding values give a mean minus one standard deviation of 1.118, as contrasted to the value of 1.24 indicated in Figure A-12. The distribution of the aged flow stress is defined by combining the distribution of the unaged flow stress (see Figure A-3) and the normal distribution shown in Figure A-13. The value of n is fixed at 6.6, as used for the unaged material, and a is related to the flow stress by Equation A-4. Figure A-14 provides various quantiles of stress-strain curves for fully saturated material as predicted by Monte Carlo simulation with 1000 trials. The stress-strain data for the saturated specimens identified in Table A-2 are also shown.

< A-16 >

0.01

0.02

0.03

0.04

0.05

Strain

FigureA-14 Predicted quantiles of stress-strainfor fully saturatedCF-8M materialat 2900C (5537F) with data points from tensile tests identified in Table A-2 The predicted yield strength can be obtained from the predicted stress-strain curve as the point at which the plastic strain is 0.002. This provides the relationship

aY

EL
aafloJ

0"002ElIee~
Eq. A- 12

A.2.2 Saturated Toughness As with the unaged material, the toughness correlations for saturated material use the Charpy impact energy. A.2.2.1 Saturated Impact Energy Distribution Figure A-15, which is copied from Figure 6 of NUREG/CR-4513 [17], shows the fully embrittled Charpy energy (Cvt) data available for CF-8M with Ni content less than 10%.

< A-17>

II

III

*I

11111

0\\ ~F-]

CF-SM
MU-

(b) .100

J rlGF

'in .,A Oi.

*-

AWH

0I 06 . 00
0 0 00

80

-040

20 IOg,o(Cv.,) =0.99 + 2.12 exp (-0.041(D))

Iog1 ,(Cv,) I
-

= 1.10 + 2.12 exp (-0.041())

-10 8

I I

II
10

I
20

i :i

:, ::I
30 40

: : : : -6
50

(]D = 6c(Ni+SI+Mn)jC+O,4NY5

Figure A- 15 Figure 6 of NUREG/CR-4513 [17] for CF-8M with nickel content less than 10% (Cw is at room temperature.) The scatter in the data is apparent. The functional form for a curve fit is given in Figure A-15 for impact energy in J/cm2 (even though the scale shown on the left of Figure A-15 is ft-lb). The corresponding function for impact energy in ft-lb is
logl0 Cvsat = 0.871 + 2.12e-. 04 10

Eq. A-13

Equation A-13 will be used to represent the median value of Cv,,, for a given (). The value of Cva for a given () is assumed to be lognormally distributed, so the probability density function of Cv,., can be expressed by Equation A-6. Equation must also be A-13 provides the median. The value of the shape parameter, Vi, defined. In the absence of data to the contrary, the value of p for the unaged material (0.279 from Section A.1.2.1) was assumed to also apply to the fully aged material. Figure A-16 provides a plot of the data from Figure A-15 along with the median value of Cv,, as defined by Equation A-13 and the predicted 10 and 90h percentiles using the assumption of lognormality with the unaged value of P. The quantiles shown in Figure A-16 are believed to provide a good description of the scatter in the Cv,, data; about half the data are above the median, with 10% of the data above and below the 9 0 " and 10t percentiles, respectively.

< A-18 >

100

80 05s
0

S60
10\

E
40

010 0=

20

30

40

50

6,(Ni+Si+Mn)2(C+O.4N)/5

Figure A- 16 Data of Figure 6 of NUREG/CR-,4513 [ 17] for CF-8M, Ni< 10%, along with median0and tted IOth and 90th percentiles Once the Charpy impact energy is known, the crack growth resistance curve is defined by Equations A-8a through A-10a. A.3 Comparisons with J-R Curve Data

In order to better understand the data and correlations suggested in NUREG/CR-4513 [17], an attempt was made to gather all saturated data for CF-8M that contain all of the necessary information. The most difficult data to obtain are the high-temperatureJ-R curve. Several NUREGs [17, 18, 21-24] were reviewed, and all of the data found for the high-temperature saturated J-R were contained in NUREG/CR-4513 [17]. Table A-4 summarizes information on the heats of CF-8M for which high-temperature J-R results were found. Tensile properties are drawn from NUREG/CR-6142 [A2], and more tensile data than what are currently in Table A-4 might be available.

< A-19 )>

Table A-4

NUREG data for CF-8M heats with fully aged, high-temperatureJ-R curves

Cast

Cent

Static

Static

Static

Static

Form NUREG Table number Cr 0 F. E 0a Mo Si Ni Mn C

N

Pipe 5385 1 17.88 3.37 0.63 8.80 0.93 0.04

-

Slab 4513 1 20.86 2.58 0.67 9.12 0.53 0.065

0.052

Slob 4513 1 19.11 2.51 0.73 9.03 0.54 0.064 0.048 4513 1 18.4 15.5 4513 1 63.1 37.2 4513 15 33 10,000 400/752 6142
3

Elbow 5385 1 17.91 3.36 0.62 8.7 0.91 0.03

-

Plate 4513 1 20.76 2.48 0.81 10.56 0.79 0.04

0.042

NUREG Table number Measured Calculated NUREG Table number

W/cm

5385 1 15.9 21.0 4513 p. A-10 113.8 67.2 4513 13 41 10,000 400/752

4513 1 27.8 24.8 4513 1 32.1 18.9 4513 15 16,20 10,000 400/752 6142
3

5385 1 19.2 24.2 4513 p. A-1 I 69.6 41.1 4513 15 36 18,000 400/752

4513 1 19.0 18.6 4513 I 30.4 17.9 4513 16 3,5 7,500 400/752

ft-lb NUREG Figure number # data Aging time Aging temperature, C/F Source
2 =

2 2 E
.21

Table number

Q.a

Yield, ksi Uhimate

30.2, 30.0 91.4,86.0 19.11 76.94 0.357 7611

0

26.7,25.0 73.8,69.2 32.48 23.62 0.326 4214 3362 18.49 83.75 0.359 7939 6188 31.92 24.02 0.327 4281 3143

(D
-a

50.81 12.86 0.311 3144 2535

C,.
m50

"' r

C. C.(0.5)'5

5942

Note 1: aging time in hours

< A-20>

Calculated 8, is obtained from composition using relationships from NUREG/CR-4513 [17]. In the derived values at the bottom of Table A-4, (D is based on measured 8,, Cvs,5o is from Equation A-13 , mso is from Equation A9a, CR5o is from Equation A-10a, and Cp~o(0.5) 5 1 is the computed median value ofJ at 0.5-in. (12.7-mm) crack extension. Charpy values are in ft-lb, andJvalues are in in-lb/in 2 (unless noted otherwise). Tensile properties for some of the heats and aging conditions in Table A-4 are available, such as from NUREG/CR-6142 [18]. Figure A-17 is a plot of saturatedJ-R data for the various heats of CF-8M at elevated temperature as reproduced from Figures 13, 15, and 16 of NUREG/CR-4513 [17]. There are 153 data points in Figure A-17. The test data fall into a fairly narrow band.
60M0

4000
C4 4

2000

0.1

0.2

0.3

0.4

0.5

Crack Extension, Inches

Figure A- 17 Saturated elevated temperatureJ-R curves for CF-BM from NUREG/CR-4513 [17) These correlations between composition, saturated Charpy, andJ-R curves are key to the statistical analysis. Figures A-18 through A-22 provide comparisons of the J-R measurements with predicted quantiles for each of the five heats included in Table A-4. Each figure also includes the lower bound as given by Equation A11 a. As noted in Table A-4, all of the heats except 205 are statically cast.

<A-21 >

0.0
-,

0.1
I
i i

0.2
. . I
i

03
. . . . I
i

II
i

.
i

.
f

.
i

,
i -i!

H - IeelO0SCF-8U29" Om 6 ' ' . . Centrifugally Cast Pipe c=21%

. ..

. 290C

. -15000

Unaged J = 650 (Aa)-' 0 i A Unaged 10,000 h at Saturatin 400TC 32 efpy at 290C J 10000

I II I

I--

"

I o I

-o -

Saturation J =328 (1A)O3"

r0
I I I III I I I I I I I I I

32 efpy at 320-C
0
I ! !

4 6 Crack Extenson, Aa (mm)

10

10000

CIA 6000

-~4000

28OO

0.5 Cak Extalon, Inche"

Figure A- 18 Test data and predicted quantiles of J-R curve for Heat 205 (see Table A-4)

< A-22 >

0,0

o~oo.10,2

0.1

Crack Ext.enson,

As (In.)
0.3

75 CF-SM Static-CaM Slab

6
A 0 Unaget 400C h at 10.000 Saturation

-25%15000
C

= 461 J g (Aa)o 37,0 Una32

"

0 0 0h0

e,,py

c at 3Jo

32 efpy at 29 0-C A AA
A

A4Y~

A --

A A A

5000

1-

188 J Saturabon 02 46

(Aa)?O3* 8 108 10

5000

4000

30

"

0.2 020.3

0.405
04

AJ-R curve for Heat quantiles af 9Crack predicted 7 and Test data Figure rA 75 (see Table A-4) of c qanil, P 19 Figure A-

Extension, inches

( A-23 >

Crack Extension, As (mm)

0.0 a a a a 0.1

0.2 I

0.3

a 290"C

Heat 74 CF4M

Stati--Cast Slab bc= 16%

-15000

Unagoed
10,000hat 400*C Saturation 32 efpy at 290'C Unaged J = 439 (.a)af"

10000 .5

32 efpyy at 320-C

00o
___
l i I i i ---4 - - 6....... .. ... ir .............. I i I

000 - ------- _ L-

00o -
J = 66 (a)(9

Saturati ,on
I a i a

.-:-'P- --

.
I= a

--

I ! I

I i I

I I

I- 0
10

0.1

0.2

0.3

0.4

0. 5

Crack Extension, Inches

Figure A-20 Test data and predicted quantiles of J-R curve for Heat 74 (see Table A-4)

< A-24 >

0.0
I
I '

0.1
I I I
I *

0.2

0.3 i
I

I
I

I .
I
I

i i

i
'

I. 290C
'

.
' -

Heat 758 CF-SM Static-Cast Elbow 6= 21%

-15000

18,000 h at 400T Saturation

~-10000
Unaged u J =487 (.Aa)O 32 efpy at 290TC
-

CE

32 etpy at 320'C

---

@"00O
" " -"I

-5000~

/
I

\
I

Saturation I
I I I

J = 244 (Aa)O
I

0 8
10

4 6 Crack Extension, As (mm)

6000

5000

40OO

C.'

3000

2000

100

0 0 0.1 0.2 0.3 0.4 0.5 Crack Extension, Inches

Figure A-21 Test data and predicted quantiles of J-R curve for Heat 758 (see Table A-4)

< A-25 >

0.0

0.1

0.2

0.3

MMa LCF-eU290 Static-Caa Plate

A 0

7,500 h6at 400 C Sataion

Unagamd

J *434 (6a)037 Saturation

3
-

2 Ofpyal'49WWM

oA, (mm) Crack ExtaaM

10

4000
90
'C

0.1

0.2

0.3

0.4

0.5

Fi9ure A-22 Test data andpredictedquantiles ofJ-R curve for Heat L (see Table A-4)

Figures A-18 through A-22 show that the test data often do not fall within the predicted quantiles. For instance, Figures A-19 and A-20 show the test data falling above the predicted 99th percentile--a highly unlikely occurrence (-~1%). All of the test data (see Figure A-17) fall into a narrower band than the quantiles shown for a given heat. In some instances, the lower bound falls at about the 1% line (heats 74 and L); at other times, it falls well below the 1% line (Heat 205 and Heat 758) and sometimes significantly above the 10% line (Heat 75).

< A-26 >

The agreement between test data and predictions in Figures A-18 through A-22 is generally poor. There are only five sets of data, whereas Figure A- 16 shows about 24 data points. More J-R data would be highly desirable. Some mixing of statically cast and centrifugally cast materials was done in the analysis previously described, but the discrepancies between predicted quantiles and test data are far beyond what could be accounted for by a 15% difference in the lower-bound expressions for these two forms of material (see Equation A-11). A.4 Material Composition and Delta Ferrite Content

Data on the delta ferrite content of CF-8M were obtained from Table B-4 of NondestructiveEvaluation:Flaw Tolerance Evaluation of Thermally Aged Cast Austenitic Stainless Steel Piping [2], which provided results for 152 heats. As described in Appendix B [1], the delta ferrite content was calculated from the material composition using the following Hull's equivalent expressions for the chromium and nickel equivalent values:
Creq = (Cr) + 1.20(Mo) + 0.48(Si)- 4.99 Nieq = (Ni) + 0.11 (Mn)- 0.0086 (Mn) 2 + 18.4 (N) + 24.5 (C) + 2.77

Eq. A-14
Eq. A-15

The composition is in wt %. The delta ferrite content (6.) is then computed from the relationship 6, = 100.3 (Creq
/Nieq) 2 _

170.72 (Crq /Nieq)

74.22

Eq. A-16

This provided 152 values of the delta ferrite for the 152 heats with available composition information. The average and standard deviation of the delta ferrite were the following: a mean delta ferrite content of 17.433% and a standard deviation of delta ferrite content of 5.384%. Figure A-23 is a cumulative probability plot of the data on normal probability scales. The data are well represented by a normal distribution.

4 A-27 >

0.999
0.99

0.9 "0.8
0.7

S0.5

0.2 0

08

0.05 0 0.2 0.01 o.005 0.001 0

10 20 30 40

delta feite

Figure A-23 Distribution of CF-8M delto ferrite from dato [A8] A.5 Degradation Parameter *

Data on the composition of CF-8M were obtained from Table B-4 of NondestructiveEvaluation:Flaw ToleranceEvaluation of Thermally Aged Cast Austenitic Stainless Steel Piping [2], which provided results for 152 heats. This was combined with the delta ferrite content from the same data set (computed as described in Section A.4) to calculate the parameter, 0, which is used to predict the value of the fully aged Charpy impact energy. The following relationship for was used: D=-S(Ni + Si + Mn)2 (C +O.4N)/ 5 Eq. A-17

The average and standard deviation of the data were as follows: mean q = 32.711 and standard deviation of D = 8.905. Figure A-24 is a normal probability plot of the data. The data are well represented by a normal distribution.

< A-28 >

0.999 0.995 0.99 0.98 0.95 0.9

0.7

0.5 0.3 E
0.1 0.05 0.02 0.01 0.005 0.001

20

,I phi

40

60

Figure A-24 Distribution of (D from EPRI data [2], CF-8M

A.6

Summary

The material presented in Appendix A provides the key ingredients of CASSPAR. The random variables in the model depicted in Figures 3-2, 3-3, and 3-4 are characterized. That is, the distribution type (normal, lognormal, and so on) for each random variable is defined and the parameters of the distribution (mean, standard deviation, and so on) evaluated. Unaged and fully saturated CF8M material is considered. The distributions are based on data provided in the references cited-primarily, NUREG/CR-6142 [18] and NUREG/CR-4513 [17]. The probabilistic model employs relationships suggested in these references, and maximum use is made of these relationships without attempting to alter the approach. The agreement between predictions and data is studied and sometimes found to be poor. The random variables are considered to be independent with no epistemic uncertainty, except for investigation of toughness-strength correlations reported in Appendix B. Information for evaluation of possible correlations was found to be minimal.

-< A-29 >

Appendix B: Toughness-Strength Correlations for Fully Aged CF-8M

The limited data on room temperature saturated Charpy energy (Cv,<) and hightemperature flow strength from the same heat of CF-8M are gathered here and analyzed to estimate possible correlation between these two random variables. A treatment of uncertainty due to the limited data is provided, which could change as more rigorous approaches are applied. Consistent with the current probabilistic treatment, the required data can be reduced to the Cv,,, and the high-temperature flow strength. Table B-1 summarizes the available data. Table B- I Summary of available data on room temperatureCwo, and high-temperature flow strength from same heats

Heat 74

37.2

2
3 4 5

Heat 75
1.2.3.10 DP2-A37
1.2.3. 12 1.2.1 DP2-A40 1.2.8.7 1.2.8 DP2-A40

18.9
63.7 10-12 Same as above

50.2 47.1 60.8 58

58.0
48.9 61.0

Same as above

4 B-1 )

Figure B-1 is a plot of the tabular data.

80

.2

40

20

20

40 Room Temperature C,., ft-lb

60

80

Figure B-I The available data on high-temperature flow strength for a given room temperature C,,o, The toughness (Cv, ) does indeed decrease with increasing strength, and a linear fit would not be unrepresentative. B.1 Least Squares Fit

Using the standard MATHCAD linear least squares curve fit for a straight line provides the following result:
Y = 8o X=Cvs,

+ ,AX
Y=aTfl

Eq. B-I

/60 = 63.284

)36= -0.283

The mean and standard deviation of the error between predicted Y for a given X and the data point are 6- = 0 and Esd = 2.965. These results can be used to predict percentiles of the distribution of a&o for a given Croa. Figure B-2 provides the 10th, 50th, and 90th percentiles as obtained in this manner.

< B-2 )

80

60

~a-.

S40

,,.

20 401

0 0

,I

,I

20

40 Room Temperature Cv,, ft-lb

60

80

Figure B-2

Available data along with selected percentiles as obtained from the least squares fit (mean)and the value of the standarddeviation of errors

The results of Figure B-2 do not include uncertainty in the values of flo, 031, and &,d due to the small set of data. Note that the distribution is quite tight (the 10th and 90th percentiles are close to one another). B.2 Predictor Intervals

The uncertainty in the predictions resulting from the small data set can be included in a couple of ways, one of which is based on procedures in Probabilityand Statistics [25]. Figure B-3 summarizes the results obtained in this way.

< B-3 )>

so 80

60

40-

20

20

40

60

80

Figure B-3 The 10th, 5oth, and 19th percentiles of distribution of aqo for a given Cv,, considering the small sample size obtainedby the Mendenhall procedure
Figure B-3 shows the effect of the small sample size in that the spread in the distribution is much larger than shown in Figure B-2. However, the approach leading to Figure B-3 does not allow separation of inherent scatter from the uncertainty due to the small sample size.

B.3

Uncertainty

It is desirable to separate out the inherent scatter (aleatoric scatter) from the

uncertainty resulting from the limited data set (epistemic scatter). This can be accomplished using procedures described in Chapter 2 of Applied Linear Regression [26]. This reference states that the beta values are normally distributed and provides formulas for values of their means and standard deviations, along with their correlation coefficient. Table B-2 summarizes the results.

SB-4 >

Table B-2 Mean, standarddeviation, and correlationcoefficient for mean line fit

13 0 01 p

62.873 -0.274 -0.872

3.239 0.091

These values were obtained using the MATHCAD program based on equations from Weinberg. Note that the mean betas are the same in Table B-1 as were obtained from MATHCAD. The standard deviations in the betas describe uncertainty in the best fit line due to the finite number of data points. The results of Table B-2 describe only the uncertainty in the best fit line (mean line); they do not include scatter about that line. The previously given values of described such scatter but do not include uncertainty from the small data set. The standard deviation of the error is taken to be itself normally distributed. The mean error is taken to be zero, and the best estimate of the standard deviation of the error is taken to be the value 2.965 previously given (,d .. = 2.965). The standard deviation of &,d is still needed. Ang and Tang (see Vol. 1, page 248, Equation 5.33 [20]) give the following expression for the variance of the variance

(s- d)
Var( d 1
Va =n-1u n

1-3 n
4 -

4 "l. Eq. B-2

where n is the number of data points (six) and #4 is the fourth central moment of the population. Using the following expression, one can estimate #4 from the data: 1

P, , ((x. -xm) 4 =_In

Eq. B-3

The xi are now the difference between the data points and the predicted mean line. The data give a value of P 4 of 208, from which the standard deviation of Fd is 2.28. Note that this is comparable to the mean value of 2.965. This large standard deviation is due to the small number of data points, and this value tends to zero as more data become available (n becomes large). Figure B-4 provides the results of a Monte Carlo simulation using the previously described approach to generate percentiles of the flow stress for a given Cv,,,. The correlation between these two variables and the small number of data points are accounted for. The Monte Carlo simulation involves sampling /J0 and /3A from their correlated normal distributions using the parameters given in Table B-1. This defines a line for each trial as defined by Equation B-1 and the sampled values of 80 and - Then, the scatter about this line is sampled. The mean error about the sampled line is zero, and the standard deviation of the error is itself a random variable with a mean of 2.965 and a standard deviation of 2.28 <B-5 )

(see the preceding paragraph). The standard deviation is sampled, and the deviation from the line (c) is then sampled with a mean of zero and the sampled standard deviation. For a set of selected values of Cv, values of O'o are calculated using Equation B-4, which is the counterpart of Equation B-1 with an error term, -.
(flo

= go +

. +6

Eq. X-4

For a given Cv,, the sampled values of arlo are sorted and the desired quantiles obtained. This provides a set of data points for a line of a given quantile, such as those displayed in Figure B-4.
80

60 90 2O

z2 0

Room Temperature C,,. ft-lb

Figure B-4 The 1Oth, 50th, and 19th percentiles of distribution of a0o for a given C, considering the small sample size using Weinberg The results of Figure B-4 are similar to those of Figure B-3, but the approach for Figure B-4 allows separation of the aleatoric and epistemic contributions to uncertainty. The band between the 10th and 90th percentiles is much wider in Figures B-3 and B-4 than in Figure B-2, which reflects the influence of the small set of data. Given that we have only six data points, the bands are not unexpectedly wide.

< B-6 >

B.4

Monte Carlo Simulation in CASSPAR

The treatment previously described can be incorporated into CASSPAR to account for the influence of strength on toughness as well as uncertainty due to the small number of data points. There will now be many runs of CASSPAR to generate a set of critical crack sizes at a given quantile. The distribution of the given quantile is then evaluated. The following steps are involved: 1. Sample flo. 2. 3. Sample/31, given the sample/lo and correlation coefficient. Sample
Esdsd

4. Enter CASSPAR. 5. 6. 7. 8. Sample Cv,,. Sample aro (using Po, P and

&sd

from Steps 1-3).

Proceed with Monte Carlo simulation in CASSPAR to generate a line describing the critical crack size for a given probability. Repeat Step 1 N, times (the number of epistemic trials).

This will generate N, lines describing the critical crack size for a given probability. The end result will look something like Figure B-5.

0.1 0.;

Confidence levels on 101 probability

O/r

Figure B-5 Depiction of the end result

The width of the band in Figure B-5 indicates the usefulness of obtaining additional data. If the width is narrow, the benefit of additional data will be low.

4 B-7 >'

Appendix C: CASSPAR 1.0 Software User's Manual

C.1 Objective
This manual is provided to guide readers in using the CASSPAR Version 1.0 software in performing probabilistic analysis on the critical flaw depth of saturated or unaged CASS at a given flaw length and failure probabilities. C.2 Introduction

CASSPAR 1.0 is an Excel' workbook-based program that works with Excel 2003 and later versions. The workbook name is CASSPAR Ver. 1.O.xls. Figure 41 portrays the procedure used in CASSPAR to define the crack depths for a given circumferential extent (0) that would fail with a given probability when the loads under consideration are applied. Figure C-1 shows the interface of the software.
~7

rvv

>v-V-rT

rs.~~i2

P. AI

I'll

_7___7_1i__:1__,1_`rW

___

---.

I I

Figure C-I Software interface

1 Excel is a registered trademark of Microsoft Corp.

< C-1

>

As displayed in Figure C-1, the software name, version number, and date are shown on the top Control Panel tab. The major interface of the software consists of three tables (Input, Input Echo, and Results), a plot, and two command buttons (Run and Re-Plot).

C.3

Required Input

This section describes the inputs required by CASSPAR for calculations and

data output. C.3.1 CalculationInputs

Figure C-2 shows the detailed information of the Input table.
13.76 13.75

210m0 10 10tw00

Figure C-2 Input tables: strength-toughnesscorrelated (a)and uncorrelated(b)

The required inputs are as follows: " " Geometry: inside radius (in. [mm]) and thickness (in. [mm]). Applied load, as follows: " Internal pressure (ksi [MPa]) Tensile force (excluding the internal pressure-induced load) (kips [N]) Bending moment (in.-kips [N-m])

Numbers of trials: when the Correlation Switch option of is set to "Yes" (as shown in Part A of Figure C-2), the numbers of aleatory and epistemic trials; when the Correlation Switch option of is set to "No" (as shown in Part B of Figure C-2), the number of trials. For the Correlation Switch (consideration of correlation between toughness and tensile properties), the options are Yes (that is, the correlation between toughness and tensile properties will be considered) and No (that is, the correlation between toughness and tensile properties will not be considered). Material type options: aged, saturated CASS material properties or aged CASS material properties will be used in the calculation.

"

"

Note that for unaged materials, the software will calculate the results with no correlation between the toughness and tensile properties, even though the Correlation Switch is set to "Yes."

<C-2 >

C.3.2 Input for Data Output

Figure C-3 shows the Results table. The software allows the user to obtain critical flaw depths for five given flaw lengths (0/n) at five user-defined failure probabilities. The user needs to input five failure probabilities in the Probability column of the Results table for each run.

1.E-05

I 0.9500

1.E-04 1.E-02 1.E-01

5.E-01

0.9500 0.9500 0.9500

0.9500

0.9500 0.9500 0.9500 0.9500 0.9500

0.9500 0.9500 0.9500 0.9500 0.9500

0.7800 0.7800 0.8400 0.9500

0.9500

0.7100 0.7100 0.7700 0.8100

0.9400

0.6700 0.6700 0.7200 0.7700

0.8100

Figure C-3 The results table with correlationoption "No" If the Correlation Switch option is set to "Yes" in the Input table, the Epistemic Quantiles table will show automatically on the right side of the Results table, as shown in Figure C-4. The user will be able to input five different epistemic probabilities for each of the aleatoric probabilities input in the Probability column of the Results table.

E0 1.E-04 10.9500

0.9500 0.9500 0.9500 10.840010.7700 0.7200 1.E-01 I0.9500 ' 0.9500 0.9500 10.950010.8100 0.7700
1.E-02
5E-01 0.9500 0.9500 10.9500 10.950010.9400 0.8100 Figure C-4 The Results table with correlationoption "Yes"

O. 0 10780010.7100 10.6700 0.9500 10.9500 10.780010.7100 0.67001

0.05

0.1J

0.5 0.9

C.4

Calculation and Results

The section describes the process of generating calculations, the input echo and the results table.

C.4.1 Calculation
After all of the required inputs are provided, the user can start the calculation by clicking on the Run button. There is no progress echo during the calculation. The workbook will be saved automatically after the calculation is finished in the folder that contains the software under the original file name.

C.4.2 Echo of Inputs

The software echoes all of the calculation inputs in the Input Echo table for each calculation, as shown in Figure C-5. -< C-3 >

13.75 Echo5

Figure C-S The Input Echo table

C.4.3 Results
11

As shown in Figure C-4, the critical flaw depth for a given flaw length (0/nr) at each of the user input failure probabilities will be displayed in the Results table after the calculation is completed. The corresponding curves Will be plotted in the chart below the Results table, as shown in Figure C-6.

0.4. Re.ul0s
0.8 --

0.6 g -e

5.E-01
1.E-01

0.2

----

I.E-04 1.E-05
, , ''

0.2

0.4 O/ar

0.6

0.81

Figure C-6 Plot of results

If the Correlation Switch is set to "No," the failure probabilities in the Results table can be modified after the calculation is finished. After changing the failure probabilities, the user can click on the Re-Plot button to check the new results without performing the calculation again if the inputs for the calculation are not changed. The numbers in the Results table and the chart shown in Figure C-6 will be updated accordingly. If the Correlation Switch is set to "Yes," the values in the Results table and the chart below the table present the critical flaw depth at the third (middle) epistemic quantile in the table on the right (shown in Figure C-4). For example,

< C-4 >

in Figure C-4, the values shown in the Results table present the critical flaw depth corresponding to the 50% percentile of the epistemic quantile shown in the table on the right. If the Correlation Switch is set to "Yes," a new worksheet called Epistemic Results will be created to show the detailed results for the epistemic band of each aleatory failure probability specified in the Probability column of the Results table (see Figure C-7).

V.

'l V.4,

V.4" U

J.+ I

.o

U.o

0.52 0.65 0.79

0.47 0.56 0.71

0.44 0.51 0.66

0.41 0.47 1 0.59

0.39 0.43 0.45

0.39 0.42 0.44

0.79
I

1 0.71 1 0.66
1 U.44

1 0.59
I U.41

I 0.54 1 0.5
I

U.UUUI

V. I OAf

0.52
0.65 0.79 0.79

1 I I i I

0.47 1 0.56 I 0.71 I 0.71 I

0.44
0.51 0.66 0.66

1 0.41

I 0.47 I 0.59
I 0.59

I .sJi I U1 0.39 I 0.39 I I 0.43 1 0.42 I I 0.45 I 0.44 I I 0.54 I 0.5 I

U.(I

I U-bW~

U.b4

I0~7 .51

I.5U.49

0.77
0.95

I0.83

0.69

0.64
0.76 0.89

0.57
0.68

0.51
0.61

0.49
0.58

0.95
0.95

L 0.95 1 0.89 I 0.77 1 0.69 I 0.65 I

1 0.951
0.95 0.95 0.95 0.95

I 0.77 I 0.69 1 0.65

1 .B 0.8 0.87 0.94 0.75 10.691 0.75 0.69 0.78_ 0.73 0.79 0.74"

U.95 0.95 0.95 0.95

1U.95 0.95 0.95 0.95

0.95

1 0.95

I 0.95
0.5 0.95.

I 0.94 1 0.79 I 0.76 I

10.95 0.95 10.81 0.81 10.771
.0.77

0J.95

0.95 0.951 0..95

0.5 .95 0.95 0.95 10.951 0.95 0.95 1 0.95 1 0.95 Figure C-7 An epistemic results worksheet

0.95 1 0.95 I 0.95

L0.82 I0.84
I

0.77 0.78 0.84 1 0.78

For the correlated results, the epistemic quantiles can be modified and replotted by clicking the Re-Plot button, without calculating the results again. However, once the calculations are completed, the aleatory probabilities cannot be changed.

< C-5 >

Export Control Restrictions Access to and use of EPRI Intellectual Property is granted with the specific understanding and requirement that responsibility for ensuring full compliance with all applicable U.S. and foreign export laws and regulations is being undertaken by you and your company. This includes an obligation to ensure that any individual receiving access hereunder who is not a U.S. citizen or permanent U.S. resident is permitted access under applicable U.S. and foreign export laws and regulations. In the event you are uncertain whether you or your company may lawfully obtain access to this EPRI Intellectual Property, you acknowledge that it is your obligation to consult with your company's legal counsel to determine whether this access is lawful. Although EPRI may make available on a case-by-case basis an informal assessment of the applicable U.S. export classification for specific EPRI Intellectual Property, you and your company acknowledge that this assessment is solely for informational purposes and not for reliance purposes. You and your company acknowledge that it is still the obligation of you and your company to make your own assessment of the applicable U.S. export classification and ensure compliance accordingly. You and your company understand and acknowledge your obligations to make a prompt report to EPRI and the appropriate authorities regarding any access to or use of EPRI Intellectual Property hereunder that may be in violation of applicable U.S. or foreign export laws or regulations.

The Electric Power Research Institute, Inc. (EPRI, www.epri.com) conducts research and development relating to the generation, delivery and use of electricity for the benefit of the public. An independent, nonprofit organization, EPRI brings together its scientists and engineers as well as experts from academia and industry to help address challenges in electricity, including reliability, efficiency, health, safety and the environment. EPRI also provides techinology, policy and economic analyses to drive long-range research and development planning, and supports research in emerging technologies. EPRI's members represent approximately 90 percent of the electricity generated and delivered in the United States, and international participation extends to more than 30 countries. EPRI's principal offices and laboratories are located in Palo Alto, Calif.; Charlotte, N.C.; Knoxville, Tenn.; and Lenox, Mass. Together.. Shaping the Future of Electricity

Programs:
Nuclear Power Nondestructive Evaluation

2012 Electric Power Research Institute (EPRI),Inc. All rights reserved. Electric Power Research Institute, EPRI, and TOGETHER...SHAPING THE FUTURE OF ELECTRICITY are registered service marks of the Electric Power Research Institute, Inc.

1024966

Electric Power Research Institute 3420 Hillview Avenue, Palo Alto, California 94304-1338 PO Box 10412, Palo Alto, California 94303-0813 USA 800.313.3774 - 650.855.2121 askepri@epri.com * www.epri.com